# Robots That Learn Surgery From Surgeons: Inner Logic's Axel Krieger

> Axel Krieger of Inner Logic and Johns Hopkins on a robot that learned a step of gallbladder surgery from surgeons, and on simulated patients.

- Show: [Audrow Nash Podcast](https://audrownashpodcast.com/)
- Episode: 26
- Published: October 6, 2026
- Length: 1 hr 36 min
- Guests: Axel Krieger (Chief Robotics Officer and co-founder, Inner Logic)
- Web page: https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c
- Video: https://www.youtube.com/watch?v=0gRJCzaB7Qo
- Apple Podcasts: https://podcasts.apple.com/us/podcast/robots-that-learn-surgery-from-surgeons-inner-logics/id1716486786?i=1000793445130
- Playlists: [Robot learning](https://audrownashpodcast.com/playlists/robot-learning-a42zvu.md)

## Show notes

Episode 26 of the Audrow Nash Podcast.

Johns Hopkins surgeons showed Axel's team the clip-and-cut phase of gallbladder surgery. The team recorded demonstrations, trained a robot on them, and it did that phase eight times in a row with no human help. The organs were from a butcher. Now they're testing it in live animals.

I talk with Axel Krieger, Chief Robotics Officer at Inner Logic and associate professor at Johns Hopkins, about why his lab moved from hand-built models to imitation learning from surgeons, how they break a procedure into small steps a robot can learn, and how simulated patients built from tens of thousands of CT scans help test surgical devices.

You'll like this interview if you're a roboticist curious about surgery, you work on imitation learning or VLAs, or you want a grounded take on when autonomy reaches the operating room. The high- and low-level policy stretch (23:46–42:28) is the most technical.

### Episode links
- Axel's LinkedIn: https://www.linkedin.com/in/axel-krieger-b14a0075/
- IMERSE Lab (Johns Hopkins): https://imerse.lcsr.jhu.edu/
- Inner Logic website: https://innerlogic.io/

## Chapters

- [0:00](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=0) Cold open
- [0:39](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=39) Axel Krieger and Inner Logic
- [4:05](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=245) How surgical robots work today
- [9:41](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=581) AI infrastructure for device makers
- [11:15](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=675) From stitching to imitation learning
- [18:23](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=1103) Autonomous gallbladder surgery steps
- [21:37](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=1297) Action chunking transformers (ACT)
- [23:46](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=1426) Breaking surgery into small steps
- [29:42](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=1782) Language between high and low level policies
- [35:44](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=2144) Adapting to surgical robot hardware
- [40:01](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=2401) How much training data it takes
- [42:28](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=2548) VLAs and large surgical datasets
- [47:30](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=2850) Mixing training data
- [51:44](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=3104) From dead tissue to live animals
- [54:47](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=3287) Self-driving analogy and FDA approval
- [58:35](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=3515) Inner Logic's data platforms
- [1:02:42](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=3762) Simulated patients from CT scans
- [1:05:20](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=3920) Simulating bones, vessels, and soft tissue
- [1:13:07](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=4387) Where surgical robotics is heading
- [1:17:28](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=4648) One surgeon, several surgeries
- [1:18:57](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=4737) Training the next surgeons
- [1:22:19](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=4939) Funding climate
- [1:24:52](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=5092) Next milestones, from suction to suturing
- [1:26:58](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=5218) Tumor resection and full procedures
- [1:29:49](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=5389) Robot surgery in 20 years?
- [1:31:15](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=5475) Advice for getting into surgical robotics
- [1:35:02](https://audrownashpodcast.com/episodes/inner-logic-axel-krieger-4hw65c?t=5702) Final takeaway

## Transcript

### [0:00] Cold open

[0:00] **Axel Krieger:** Ultimately, we're able to do eight of those back to back without needing any human intervention. So, We have fewer and fewer surgeons available. We have an aging society. We have a gap of in the next 10 years we are missing 20,000 surgeons, My vision was eventually you might have a surgeon supervising several procedures at the same time.

[0:25] **Audrow Nash:** Yeah, yeah like a grocery store checkout model for surgeons. What a world. would you be surprised if it wasn't the case in 20 years from now? That you could just like go and have a robot operate on you?

[0:29] **Axel Krieger:** I looking at the trajectory and being in it, I would be surprised

### [0:39] Axel Krieger and Inner Logic

[0:42] **Audrow Nash:** So, hey Axel, would you introduce yourself?

[0:45] **Axel Krieger:** Hello, and thank you so much for having me. I'm Axel Krieger. I'm the chief robotics officer for Inner Logic. I'm also associate professor at Johns Hopkins where I'm currently on sabbatical.

[1:00] **Audrow Nash:** Tell me a bit about Inner Logic and how it relates to what you're doing at Johns Hopkins.

[1:07] **Axel Krieger:** Fantastic, so my research is on autonomous robotic surgery. So currently robots that I used in the clinic are operated every step of the way by an operating surgeon. So surgeon sits on a console, controls every step. And my vision is really that you know robots can perform some critical steps autonomously and surgeon takes on supervisory role. And so we feel this is coming, autonomous surgery. And so we founded the company Inner Logic with Tito and Matthias, a neurosurgeon Tito who also has an MBA. So amazing combination. And Matthias, a fellow professor from Hopkins, also a fellow German American. And he's a machine learning AI expert. And so that combination started Inner Logic and we believe autonomy is coming. And we want to provide companies with the infrastructure to help them on the way to that. So we support different medical device manufacturers with data models, functions on their ambitious AI roadmaps.

[2:36] **Audrow Nash:** Very very cool. Tell me how, like tell me more about Inner Logic, kind of just the logistics about it. Like I believe you guys have started pretty recently, tell me how big you are, tell me about the funding you guys have so far, all the kind of like logistics of the company.

[2:53] **Axel Krieger:** Yeah, we started fundraising last fall. Kind of interesting, we received the first outright from a VC after publishing a science robotics study on autonomous cholecystectomy. And that kind of really spurred us. We were already talking in the background about maybe doing something together. And that really got us into gear. We then did fundraising in the fall, raised 11 and a half million dollars a seed by Collet, by General Catalyst and Bison. And then really became operational around January. So moved into our new space in February and hired our first five founding engineers and are now already close to 20 people down in Baltimore and hiring also worldwide and more remote. And so yeah, and signed our first contract with device manufacturers. So exciting, exciting whirlwind last year.

### [4:05] How surgical robots work today

[4:05] **Audrow Nash:** Yeah, really. Okay, hell yeah. Tell me, so for context, tell me where the robotic surgery space is. Like I like to understand where we're coming from and then kind of why you guys are doing what you're doing.

[4:20] **Axel Krieger:** Yeah, so a bit of context, right? So, you know, surgical robotics has enabled you know the shift from open to minimally invasive surgery. So, you know, we used to have, you know, large, you know, scars opening, having direct access with your eyes, hands on organs. And what robots have allowed is that you can sit on a console and drive these, you know, keyhole port based surgeries with dexterous tools and perform these, you know, complex, you know, surgeries now minimally invasive. And really the main focus over the last few decades has been to make the robots smaller, you know, more dexterous allowing, you know, natural orifice or access, for example, through your mouth and other, you know, orifices, you know, to get to the organ of interest without, you know, much bleeding, scarring. So, you know, less invasive. And, you know, that has been, you know, the progression in technology over the last two tech you know decades. And over the last few years, and now I think really projecting into the future has been the shift with more software centric, more, you know, intelligence, smarter, you know, tools and devices to, you know, reduce the burden of assertions and, you know, really help our patients to, you know, have fewer complications and better outcomes. So, it's been a shift of mechanical innovations to more software innovations and intelligence and autonomy. And, you know, we've been working on this, you know, since about 2012 on autonomous surgery that was, you know, considered still a little crazy at the time. And now, you know, I think really that the consensus is that, you know, AI in surgery is coming, is helping. And, you know, we are really helping at, you know, logic, you know, that development cycle, you know, to accelerate and bring, you know, forward those smarter intelligence, more autonomous devices.

[6:41] **Audrow Nash:** So just so I can picture it a bit more clearly, you said like keyhole and port. I've seen some of the DaVinci ones and I've tried them at different conferences. Not, tell me if this is a good representation of more or less what surgery robots are like. It was like I was looking in a screen and I had my thumb and pointer finger controlling little robot hands that were like on the ends of sticks. So it was like a little claw on the end of a stick and I would control that and you could stack blocks or sort rings or whatever. That's pretty much what robot surgery means at this point, focusing on hardware. Is it correct and like the problems have been like making those hands as small as possible so it could be minimally invasive and then maybe providing some tactile feedback to the user so that they could feel if they were gripping something or this kind of thing. Is that about right or what do you think? So

[7:33] **Axel Krieger:** Yeah, really, really good summary, right? So, the operating surgeon sits on a console, puts their head into like this, you know, camera, and then you get like a 3D vision of the scene because you have 3D laparoscopes. And then you have two hand controllers. Robots typically have four arms that you can switch to, which one you're using. And then you control kind of two arms at the same time. And it becomes really an extension of the surgeon, right? Like, you know, that you can work inside the body. You inflate it with CO2 to get like some distance and some room to operate in. And then you can perform this. But every motion of the robot is, you

[7:57] **Audrow Nash:** Cool. Hmm.

[8:17] **Axel Krieger:** Know, exactly, you know, just mimicked what the surgeon does on the console to what the, you know, robotic system does over the patient, sitting in the same room. And, you know, currently there's really no autonomy in these procedures. That's the standard. So, for the last, you know, decades, the robots have gotten better control, more surgical tools, you know, more staplers, more energy devices, you know, smaller,

[8:44] **Audrow Nash:** Cool.

[8:45] **Axel Krieger:** You know, more stable, less, you know, vibration, you know, and making them more flexible to enable some procedures that are, you know, even further away from like a laparoscopic entrance, you know, like more like snake-like robots. So, a lot of development on the mechanical side. And really also like an explosion of different companies coming in, you know, Intuitive is the main, you know, robotic surgical company that, you know, a lot of people know the Intuitive, you know, DaVinci robot. But now we have, you know, you know, J&J robot, Metronic robot, you know, Moon Surgical, CMR, you know, some Chinese companies. So, you know, I think around 60, you know, surgical robotics companies are, you know, developing specialty robots for all different organs and, you know, bringing those to market.

[9:25] **Audrow Nash:** Wow. Okay. And so most of these companies have been innovating on the hardware side and you guys, you mentioned infrastructure is where you're interested in. Tell me a bit about that and tell me how that works. It's like you won't be making the hardware but you'll be working with all of the hardware that companies make. Tell me about that.

### [9:41] AI infrastructure for device makers

[10:03] **Axel Krieger:** Yeah, exactly. So, a lot of the, you know, companies' expertise, core expertise, is more on the mechanical side. And, you know, and you know we really see a shift there in the focus and, you know, like the roadmaps are, you know, quite ambitious on, you know, really improving the, you know, guidance, you know, decision-making, you know, control. And we help companies on their roadmap. So, you know, how does it look like? So, let's say you want to, you know, automate a suction functionality, right? So then we help you with, you know, getting data, with tool tracking. We can, you know, help with, you know, foundation models to help you with, you know, learning those functions faster. We help you with testing. So, and all the way to, you know, really, you know, co-developing, you know, functions. So, it starts often with, you know, data needs, but then, you know, really, you know, working on expanding this to more and more functions.

### [11:15] From stitching to imitation learning

[11:15] **Audrow Nash:** Mm. That's really cool. And where, tell me about the science robotics paper you all published and how that kind of kicked things off and where you think we're going.

[11:29] **Axel Krieger:** Yeah, maybe I can, you know, explain this, you know, in context to kind of our research, right? So, we started, you know, 2012 in autonomous robotics and started, you know, with a traditional model-based approach, right? So, you know, separate perception tracking, you know, modeling the interaction between tool and tissue and then, you know, controlling that, you know, and making adjustments. And so, we got pretty good in suturing, you know, with specialized camera, specialized robotic tool, you know.

[12:02] **Audrow Nash:** Now, I don't really know medical terms and I'm supposing people do not either. Suturing, does that mean sewing up a wound or a, okay.

[12:10] **Axel Krieger:** Exactly, exactly, exactly. So, you know, suturing is, you know, where you have a suture needle and you need to sew up a wound. And what we've been working on, yeah, like they're like kind of half circles, you know, half circles.

[12:16] **Audrow Nash:** Mm-hmm. Like one of those hook shaped needles.

[12:25] **Axel Krieger:** And, you know, the particular application that we worked on was on the bowel anastomosis is called the closing of a bowel. So, imagine you have a tumor in the bowel, you take out a section, right, and then you need to put it back together. And putting that back together requires placing, you know, 20 little stitches around the bowel. And if you miss one little, you know, mark, that's where you have a leak, right? And so, you need a lot of consistency and repeatability. And so, that's where we started, you know, technically building a robot for this anastomosis. And, you know, we were able to show that we can perform this in an animal model, you know, very consistency in a supervised, you know, autonomous function. So, you know, surgeons don't have to, you know, prescribe every motion of the robot anymore. The surgical system, you know, tells you where it wants to go. You can make adjustments. And every now and then it misses a stitch and you provide a little bit of help. So, that's, you know, how we, you know, traditionally build robots. And, you know, it, we hit a bit of a ceiling, right? So, about 70 percent of stitches, you know, were autonomous, 30 percent needed help. And we, you know, really learned how to do stitching, but it didn't translate to other procedures, right? And so, you know, where we pivoted then, you know, a few years ago is really to, you know, a fully, you know, end-to-end learning-based approach,

[13:49] **Audrow Nash:** Mm-hmm.

[14:04] **Axel Krieger:** Where the idea is that we can use imitation learning to, you know, teach our robots how to perform these procedures. And what is exciting about imitation learning for surgery is really we have this large install base of, you know, surgical robots. And we have thousands of incredibly highly trained surgeons that are, you know, in their residency and fellowship training. They are already using the robot. So, they grow up with the robot. They are, you know, deep, deep, deep specialists on how to use, you know, robotic technology. And so, I mean, if you watch some of the surgical robotics videos, how they perform surgery, it's just beautiful. I mean, it's like an art how every motion is just so,

[14:27] **Audrow Nash:** Mm-hmm. Hmm.

[14:56] **Axel Krieger:** You know, you know, exquisitely, you know, detailed, great visualization. I mean, they're really, really, really good at this, right? And so, instead of modeling this and really trying to automate this, you know, with, you know, a rule-based, you know, set up the idea here is where we pivot it to is that we can learn directly from those expert surgeons. And I just feel like for the entire robotics community, like this, you know, field is so interesting because, you know, what other field has, you know, so many expert robotic users, right? Like if we think about, you know, folding laundry, right? Like everybody has to kind of learn it, right? But in surgical robotics, we have that pool of, you know, fully trained, you know, surgical robotic, you know, surgeons. So, that's kind of like to, you know, preface the kind of switch and the history of our development in the lab a little bit.

[15:54] **Audrow Nash:** Mm-hmm. I have one question with all this. With the imitation learning from the robot platforms,

[15:55] **Axel Krieger:** And, yeah.

[16:02] **Audrow Nash:** Do they do very good logging? So you end up having a very good ability to replay everything. You see what the cameras saw. You see all of the forces that the surgeon is putting on things. Maybe you can even have like a transcript of it for context or things like this. Like do you have a very good data capture of surgeons from them using the robot systems and does it vary quite a bit between the different companies and their different robots, or how is it?

[16:31] **Axel Krieger:** Yeah. So, typically, the companies, you know, really collect and log the data well, right? And so, you know, what's very common is that surgeons capture the video and, you know, so video data is available a lot. And companies collect, you know, also the kinematics data. So, the robot, movement data, right? And so, what is difficult as a researcher is getting access to both, right? And so, that's definitely a hurdle for us. Yeah. And also like

[17:03] **Audrow Nash:** Yeah, yeah. Part of the reason to start the company in some sense, because then you can have collaborations directly.

[17:13] **Axel Krieger:** HIPAA rules, right? So, this is for patient privacy, right? So, it's of course patient data, right? So, sharing that is not you know, super easy. So, there's definitely hurdles around, you know, the collection of that data set. But companies have access to it and there are also ways of, you know, really consenting patients and getting this, you know, ethically approved to share. Because, I mean, you know, it's not like patient identifying. We cannot really trace that back to a particular patient, you know. Yeah.

[17:14] **Audrow Nash:** Oh, yeah. Yep. Yeah. Mm-hmm. Yeah. So cool. Okay, so that was the context. And then moving into the next part.

[17:55] **Axel Krieger:** Yeah. So, that's the context. Why is imitation learning so exciting? And so, we started, you know, exploring imitation learning, you know, by, you know, just figuring out how to perform, you know, small different surgical tasks, you know, like just grabbing a needle. And then, you know, in this paper that we published last year, you know, really tied together small little surgical steps to perform a complete phase of surgery, you know, autonomously. We picked the gallbladder surgery here. Gallbladder surgery is, was the first that was converted from open surgery to laparoscopic. So, it has a bit of this lighthouse character of being the first, you know, and so, it was a good target for us.

### [18:23] Autonomous gallbladder surgery steps

[18:31] **Audrow Nash:** Mm-hmm. Cool. Why is it that way? Is it a simpler surgery, or does it lend itself very well to laparoscopic, which means like the little hands going in, I guess, right? Little holes.

[18:54] **Axel Krieger:** Exactly. That's the little keyhole surgery. Yeah. Yeah. It's, you know, the it is a, you know, very common surgery. So, it's done a lot and it was, you know, it's, you know, well accessible, right? And so, it is,

[18:56] **Audrow Nash:** Mm-hmm. Is it like removing gallstones, or what's the point of it?

[19:15] **Axel Krieger:** It's removing the gallbladder, right? If you have a gallbladder infection, you remove the entire gallbladder.

[19:17] **Audrow Nash:** Oh, removing it. Ah, okay.

[19:21] **Axel Krieger:** And so, you know, there's three phases to this. There's, you have to dissect the gallbladder from the liver bed. Then, you have to find and isolate the little artery that supplies the gallbladder and the little duct that connects the gallbladder to the bowel. And those two structures, you have to place like small little plastic clips in it, three clips, and then you have to cut in between.

[19:37] **Audrow Nash:** Oh, like clamp it or something. Mm-hmm.

[19:50] **Axel Krieger:** And then you take it out. So, that's kind of like, you know, in simple words, how to perform that surgery. And we focused in our autonomy work to make it a bit easier, not on the entire gallbladder surgery, but this kind of middle phase of finding the duct and artery and then placing clips and then cutting it. So, that's kind of the clipping and cutting phase of the gallbladder surgery. And so, how did we, you know, perform this? We found our, you know, clinical collaborators at Johns Hopkins to show us how to perform that. So, they came to our lab, set up you know a ex vivo, that means, you know, just the parts from a butcher, the gallbladder and the liver. So, we got those you know, in from butchers online.

[20:45] **Audrow Nash:** Huh. It's like a cow, or I guess,

[20:50] **Axel Krieger:** Yeah, from pigs, from pigs.

[20:52] **Audrow Nash:** For pigs. Oh, yeah, yeah.

[20:53] **Axel Krieger:** They're quite similar to human organs, right? So, we could order them and they showed us how to perform it, right? How to, you know, isolate it, you know, visualize it, and then perform those steps. And our engineers then, you know, and students really learned how to perform it. And then we collected, you know, data for the different steps of that clipping and cutting and trained, you know, then policies to perform that and got better and better. And then, ultimately, we're able to do eight of those back to back without needing any human intervention. So, we didn't need any changes

[20:58] **Audrow Nash:** Cool.

[21:34] **Axel Krieger:** And it was able to perform all of eight of them successfully.

### [21:37] Action chunking transformers (ACT)

[21:38] **Audrow Nash:** Hell, yeah.

[21:38] **Axel Krieger:** And that's kind of like the study that we published in Science Robotics.

[21:42] **Audrow Nash:** Okay, and what was the approach that you guys used? How did you... Like, so I feel like I haven't been too up-to-date with it, but it seems like imitation learning is really becoming a good way of teaching robots to do things. Like, there's been a big resurgence in interest in there. Tell me what's changed, and why all of a sudden we're seeing all these really good imitation learning results.

[22:13] **Axel Krieger:** Yes, so the backbone is an imitation learning. So, there we used, you know, in collaboration with Chelsea Finn in Stanford, you know, really basing a lot on her work in the ACT policy. So, that's the action chunking transformer,

[22:17] **Audrow Nash:** Mm-hmm.

[22:32] **Axel Krieger:** You know, that was really deployed there for more industrial-type robots. And, you know, we used that as kind of like the backbone for our imitation learning. So, I think, you know, big, you know, shift was, yeah.

[22:47] **Audrow Nash:** Can you tell me, like, at a high level how it works with This ACT, chunking?

[22:51] **Axel Krieger:** How do you train a robot? Yeah, so imagine if you train a transformer model to find the next little robot action step. And you train it by, you know, pairing videos with how an expert surgeon moves the robot to imitate that, right? So, that's kind of like, you know, the basic principle. So, the, you know, system looks at, you know, the current video, has, you know, is trained on, you know, all the different actions that an expert did and finds the nice, you know, the next little robot action to get closer to the goal. And so, that's kind of like a little chunk of action. And so, it performs that and then find the next action, the next action, the next action until you can like piece that together to perform, you know, the entire procedure.

### [23:46] Breaking surgery into small steps

[23:47] **Audrow Nash:** Mm-hmm. Achieve it. How do you segment actions for this? Is it like a new input, or... How do we do the action segmentation? Because I imagine it could be, like, infinitesimally small actions, or large, over big periods of time, actions. How do you break them down?

[24:08] **Axel Krieger:** Yeah, that's really, really the trick to make this work, right? If you try to automate a long horizon, so a really, you know, long, you know, action sequence, then you have a lot of failures and problems. And so, you know, parsing it into like the smaller fundamental steps is the key, right? And so, what we deployed there is a hierarchical,

[24:17] **Audrow Nash:** Too much.

[24:36] **Axel Krieger:** You know, policy that, you know, first use, you know, segment the, you know, clipping, cutting phase of this cholocystectomy into small little fundamental steps using like a surgical taxonomy, like, you know, like what are the fundamental steps of the surgical procedures, you know, like, you know, clipping the first clip into the, you know, artery, like small little, you know, sub-steps. And then in deployment, right, you run this hierarchical model that first what's the correct next step and then executes it under low-level policy. So, that's how you know, have a pretty simple policy achieving, you know, complex, you know, procedures,

[25:28] **Audrow Nash:** Behavior.

[25:30] **Axel Krieger:** You know, step by step in a, you know, like a surgical, you know, segmentic way.

[25:36] **Audrow Nash:** Mm-hmm. Do you define the actions, the breakdown of the actions manually, using that, like, surgical tectonomy? So are you teaching it, "This is the first part, this is the second part"? Or is it learned implicitly through watching? Or how does that work? And if you are doing it manually, it sounds like you probably have to iterate a little bit and try to figure out how big or small to break down the steps to make them reasonable for the robot. Like, I'm sure sometimes they don't perform as well if you're like, "Well, do the thing." It's like, it's too big. If you, say, do this one clip, maybe that works better. But, or as opposed to do these clips, this kind of thing. How do you break that down? Is it manual? It looks like you're going to say it's manual. But yeah, go ahead.

[26:21] **Axel Krieger:** Yeah, you can foresee my answer already. Yeah, it was very manual and absolutely trial and error. And sometimes the step was too large and then we had to really break it down further. Yeah.

[26:29] **Audrow Nash:** Mm-hmm. Huh. Tell me about that process. Does that mean what you do is you try to take the biggest step and then you see bad performance and then you make it smaller and you just keep doing that until you arrive at a good working thing? Or how does, how do we achieve these steps?

[26:48] **Axel Krieger:** So, of course we started with just teaching small little steps on its own. And then we built up an intuition, what's the right kind of step size? And so then looking at the procedure, we were able to already kind of like, really have a good idea what could work. And then if we then have one step and struggle really with it, then we often further divide it into more. So that's kind of like how it works. You build an intuition, you do an initial test run. And then if you really struggle with a certain procedure, either you did something wrong in data collection, you need some recovery data. Or you know or this or it's just too ambitious, you cannot do it, then you subdivide it. Yeah.

[27:18] **Audrow Nash:** Subdivided. So, what does it look like to subdivide the data for this? Is it like you're segmenting the video? Would you say, here to here, this is the step. Here to here, this is the, that is what it is? Okay, very interesting.

[27:52] **Axel Krieger:** Yes, yeah, so the subdivision can really happen on the video. And the video, you know. and you know the you know high level policy that selects what is the current next step, really runs off of video as well, yeah.

[28:09] **Audrow Nash:** Very interesting. So, I suppose what you become good at is you become experts at to different problems. Is that kind of like, so first it's the gallbladder surgery and then I suspect there are other areas you're focusing on too. But you basically become very good at breaking down the problem for the robot to learn and then do. Is that kind of, is that a fair way to think of this or what do you think?

[28:40] **Axel Krieger:** Yeah, I think that's definitely a big, you know a important part of it. The nice thing is that surgeons have worked on this idea of the language of surgery. And also scientists quite a lot already, before

[28:54] **Audrow Nash:** Quite a lot, probably. Mm-hmm.

[28:58] **Axel Krieger:** Imitation learning came along, right? So there's a body of work already in this taxonomy and like what are small surgical steps. There has been some cool work recently from Andrew Hung, that correlates these different steps and the sequence of steps to outcome. to you know like you know So that's been really nice where we can kind of piggyback on body of research

[29:18] **Audrow Nash:** Mm-hmm. All that existing work. Mm-hmm.

[29:26] **Axel Krieger:** Already. And what we've seen is that the smallest kind of, it's called a gesture, that's the surgical term for the smallest action. That is actually also a really nice and small step that we can use for imitation learning as well.

[29:35] **Audrow Nash:** Yeah. Cool. Do you use, so tell me the role of language in this. How does, how does language relate to the video segments that you're doing? So it's basically you labeling what the segment is and that's used to, I want to understand how much of the structure is given by a human and how much of it is kind of learned in how to do the sequence or this kind of thing. And then also how do the words relate to the video segments for this kind of thing? Are we, are we using LLMs in this? Is it a visual language model or like, how does all this work?

### [29:42] Language between high and low level policies

[30:22] **Axel Krieger:** Yeah, wonderful. So yeah, language, we use language to connect the high level policy from the low level policy. So how they interact is through language. And what is cool about language is that it also mimics of course, how a surgeon would you know, oper operate together with a trainee, right? If you have a you know n you know medical student trying to do a procedure the first time with experienced surgeons, how do they interact with language, right? And so if the trainee says, I wanna do this next step, right? Then the supervising surgeon can say, no, no, do something else. So it's really nice from the concept of supervision.

[30:47] **Audrow Nash:** Mm-hmm. Oh cool. It's a way of doing evaluation probably too. Is it? Because you can be like, oh it's suggesting do this and no, no, no I don't want to go that path yet. This kind of thing.

[31:19] **Axel Krieger:** Exactly, that supervision is kind of built into the structure and then changing it and then improving the learning by learning from a wrong labeling and then getting it right. So you know, so that's why this is good and how it works technically is you manually label the different steps and then train a network to figure out what the correct step is based on the history of video that you give. So in this instance, we gave it four

[31:43] **Audrow Nash:** Segments. Mm-hmm.

[31:56] **Axel Krieger:** Images, four seconds apart, and then

[31:59] **Audrow Nash:** Aww.

[32:01] **Axel Krieger:** Train the network to figure out what is the next step that we need to perform. And that is the basic part, it's just that language that is labeled and then we added a correction part, right? So the second functionality of the high level policy is really figuring out if everything is going well or if the low level policy is struggling or needs some correction. So that's more like a flag and then a directional flag, like you need to go a little bit more left, up, down, and

[32:30] **Audrow Nash:** You have an evaluator loop. Yeah, yeah.

[32:43] **Axel Krieger:** Then the low level policy takes that step and takes, do I need to do a correction? And you need to change a direction in a certain way as a language conditioned low level policy that then takes this into account by then producing the current robot action and that next little chunk of action that we need to produce.

[33:10] **Audrow Nash:** Cool. Okay, I don't know that I fully understand all of it, but the thing that I'm starting to see is the challenge with LLMs now is they can't ingest video. And so part of what you're doing is you take a series of frames, you learn how to talk about it like a surgeon would, and then you use that to understand your current state and to pick what you're doing in the future for this kind of thing. Is that right?

[33:39] **Axel Krieger:** Yep, yeah. It's you know, Like very similar to a vision language model, yes.

[33:40] **Audrow Nash:** Super cool. Uh-huh. What's the, how, so when we're saying high level and low level, what exactly do each of those mean? Is the low level like literal joint commands for the robot, or is it like actions that we're defining for the robot? And then the high level is like keeping track of the, what we're actually doing for the surgery and how the progression of steps. And then how, is there like just a language model running on this where it's trying to get through like a markdown task list of all the things to do for a surgery, or how does this actually look?

[34:22] **Axel Krieger:** Yeah, so the high level has this input history of video and as output, just language, nothing else. you know? And the low level has as input the latest video frame, the language command of the high level policy, and then the output of the low level is the action of the robot. So that's kind of like you know, it's kind of separating the architecture into one kind of decision making step and then into an action step. So it makes it easier to

[34:30] **Audrow Nash:** Cool. Mmm.

[34:56] **Axel Krieger:** Train each of them separately,

[34:58] **Audrow Nash:** So cool.

[34:59] **Axel Krieger:** Makes it easy to supervise, know what's going on, and really just was very successful for us to putting this together.

[35:09] **Audrow Nash:** Mm-hmm. That's such a cool way to do it. Was there, so is this very similar to This ACT framework that Chelsea Finn and the groups over there have worked on? Is that pretty much the same thing, or how is it similar or different?

[35:19] **Axel Krieger:** Yeah, the ACT framework is the backbone to our low level policy. So that's basically taking the latest villa you know video in and video framing and producing the action. That backbone is ACT, small variation there, and then also really adapting it to surgical robots. Surgical robots are

[35:31] **Audrow Nash:** Mm-hmm. Yeah.

### [35:44] Adapting to surgical robot hardware

[35:49] **Axel Krieger:** These long stick tools. They have a lot of deflection. They have cable drives. They are not repeatable as an industrial robot. So a lot of our key innovations was to adapt this, you know, to come up with the right action representation to you know really make this work on these more cable driven and more flimsy surgical robots compared to really well encoded industrial robot arms.

[36:20] **Audrow Nash:** Industrial arms. Huh. Very cool. Okay, so the actual tool that you're using is different, and that makes it harder because industrial arms are very stiff and these ones flex a little bit that you use in surgery. Were there any other, I was just going to ask, differences between like what you were able to get off the shelf in some sense and using it For surgery, like what were some of those challenging aspects? So obviously stiff versus kind of soft actuators, or not soft, but less rigid, say. Any other major difficulties in applying kind of like an industrial solution to a surgical application?

[37:07] **Axel Krieger:** Yeah, surgical robots have replaceable tools, right? So our surgical tools are kind of rated to being used maybe like 10 times and can be sterilized 10 times and then they are thrown away. So they really wear out quickly, right? And so, you know, you know struggle was, for example, you train on one tool and then you replace it and have slightly different kinematics and then struggle again, right? So that is certainly a problem. The use as hardware, the DaVinci research kit. So these are older generation DaVinci robots where the company donates the hardware. And together with Hopkins, you know, develop here at Hopkins and WPI is to you know ha build a controller infrastructure that we can then you know use the hardware without exposing the software layer and the control layer of the original intuitive DaVinci robot. So it is basically replacing all the electronics and software layer. So that is the robotic test that we are using.

[38:19] **Audrow Nash:** Mm-hmm.

[38:24] **Axel Krieger:** There are different generations of it. And so another struggle for us was we you know we recently started doing you know in vivo, so living animal tests on a different generation of this robot, and there we also saw large kinematic differences and our policies did not just work out of the box. So those are kind of the struggles.

[38:43] **Audrow Nash:** Yeah. Yeah. Different data distributions. It's always hard when you change hardware for everything. How did you resolve that? Like, did you just have to retrain on quite a bit of new data? Or what did you do when you got different hardware?

[39:03] **Axel Krieger:** Yeah, that's been kind of like the solution is to, yeah, exactly. You add more data with a new configuration and then can combine it at least to get best performance. Yeah,

[39:05] **Audrow Nash:** Because that's a common problem. Yeah. Do you do a lot of like the, I don't know, maybe something that I've seen work a little bit is you can try to leverage the existing data as much as possible, but then you fine-tune on some of the newer hardware to make it work? I don't know. Is that kind of the approach that you've taken, or what did you do?

[39:35] **Axel Krieger:** This kind of act backbone is a little bit small in parameter size, and so it doesn't lend itself to fine tuning. you know, We are doing fine tuning a lot.

[39:39] **Audrow Nash:** Yeah. Oh. You just restarted, I guess.

[39:47] **Axel Krieger:** Yeah, I mean, you just kind of co-train it with the new robot data, so that's been the solution.

[39:54] **Audrow Nash:** How much training is required for it? So you were working, the science paper was step two of three, of three steps in the gallbladder surgery. How much training data do you need to be able to do that reliably? So you said, less than ten cases successful all in a row. How much training data did you need for that?

### [40:01] How much training data it takes

[40:18] **Axel Krieger:** We had about 17 hours of training data, so that's the order of magnitude. We

[40:24] **Audrow Nash:** How long is one surgery? Because I mean that would be, like how many instances of gallbladder step two did you end up having?

[40:30] **Axel Krieger:** Had 30 different gallbladders and performed each of the steps about 15 times. So it's like 30 variations of gallbladders, but on each of them we performed kind of like each step 15 times to get us to around this 400 or 500

[40:49] **Audrow Nash:** Yep. Instances of doing it.

[40:57] **Axel Krieger:** Instances. Yeah.

[40:58] **Audrow Nash:** Okay. And that was, what's, I don't know if you probably have an idea for this, I don't know if this is the only question. How was the performance over time as you were getting those samples? Like at 100, 200, 300, 400, 500, if you did 10,000, would it be unbelievably more? Like it probably asymptotes in performance at some point. How's that, what's your mental model of how that's happening with your data?

[41:29] **Axel Krieger:** Yeah, we definitely study that quite a bit to figure out what's the right you know amount that we need.

[41:36] **Audrow Nash:** Number to go to.

[41:38] **Axel Krieger:** So it really depends on the complexity of the procedure and the manipulation task, but it's typically around 200 to 500 episodes is where we get some good performance there, and then it does asymptote. So we get small improvements, but not massive improvements afterwards with these smaller models. We have, of course, since the science paper come out, we're also working a lot with larger VLA architectures that are more parameters that also train on

[42:16] **Audrow Nash:** More than the ACT model, you're saying.

[42:19] **Axel Krieger:** A variety of surgical robotic procedures and then fine-tune on a particular case, like you mentioned. So we've been, of course, working on these as well. Yep.

[42:25] **Audrow Nash:** Yes. Can you tell me a bit, I have been out of the loop a little bit. Can you tell me kind of fundamentally what visual action language models are? Like what, how is the approach, I guess ACT is a much smaller model for learning things but these are more flexible because they're bigger and I'm curious how you use an LLM And like what it actually looks like in implementation.

### [42:28] VLAs and large surgical datasets

[42:52] **Axel Krieger:** Yeah. So I think, you know, kind of simply described is, you know, relative, you know, similar between, you know, like act as a small model and then, you know, the, you know, larger VLA, a lot more parameters and, you know, a kind of like, you know, natural, you know, you don't have to like specifically separate in this high level and low level policy. That's kind of like integrated in it, you know, but, you know, from a functionality, video input, action output, and, you know, language to, you know, give you that context, you know, what are you performing?

[43:21] **Audrow Nash:** All the steps. Oh yeah, yeah. Is it like a more general version of what you were describing with the high and low level for this? Does it draw its own abstractions or does it, how do you, I guess my assumption with this that you're describing is you end up needing a lot more data for it to be performing. Is that correct? But you can do the fine tuning and stuff so maybe that's the direction you're going for it being more robust on various hardware because you're aiming for general infrastructure for several companies. Is that correct?

[44:09] **Axel Krieger:** Yeah. You're really correct. So you have, you know, the more capabilities, also more training requirements, but then the ability to, you know, generalize it and then, you know, fine-tune it on a specific, you know,

[44:21] **Audrow Nash:** Generalize is much higher.

[44:27] **Axel Krieger:** Work. So, you know, to give you a good example of this, we started together with NVIDIA and 50 different institutions, this push for robotics data sets with video and kinematics. And so we received, you know, there were eight companies and involved giving us also some clinical data like CMR Surgical, for example, delivered, you know, hundreds of hours

[44:42] **Audrow Nash:** Cool.

[44:54] **Axel Krieger:** And then, you know, 50 different institutions. And so we have, you know, this large data set that is, you know, very diverse, all different robots,

[44:55] **Audrow Nash:** It's amazing.

[45:06] **Axel Krieger:** And there we can, you know, train, you know, large reveal architecture on, you know, general surgical robotics tasks and then fine-tune them on, you know, specific robots and specific application tasks, you know.

[45:22] **Audrow Nash:** So do you think the core competence of Inner Logic is going to be effectively you'll have your visual language action model that's like the core thing and then you'll fine tune it to applications and procedures. Is that kind of the play you think or how do you think about it?

[45:43] **Axel Krieger:** So, you know, we, you know, this imitation learning, you know, framework is kind of one, you know, direction that we are exploring, right? So, you know, we are really supporting, you know, procedural,

[45:50] **Audrow Nash:** One aspect.

[45:56] **Axel Krieger:** You know, device manufacturers in a much broader sense, right? And so it's not all, you know, action-related. So, of course, think about, right, you know, like surgical robotics, you know, kind of the last aspects what on the roadmap is kind of the autonomous action, right? So, you know, it's a lot in the analysis side, it's a lot in the guidance side, in the decision-making,

[46:04] **Audrow Nash:** Yeah. Autonomy.

[46:21] **Axel Krieger:** You know, and the action is kind of the last piece, you know, a piece. And so what we've been really finding is, you know, a lot of needs already in the kind of earlier phases. So, you know, we are certainly, you know, we have our, you know, research in VLA and world models, you know,

[46:41] **Audrow Nash:** Yep. That's kind of the end game in some sense but there's a whole bunch of things leading up to there that are all very promising and needed. Yeah because I mean guidance would be really cool.

[46:44] **Axel Krieger:** We... Exactly, exactly, exactly, yep.

[46:52] **Audrow Nash:** Like there's all sorts of things if you could have a more junior surgeon doing, being advised by an AI that's like a senior surgeon that would be super cool. Especially in like maybe crisis moments or anything else too.

[47:06] **Axel Krieger:** Absolutely. So, an anatomical, you know, segmentation, like finding where's the artery, right? Like avoid that artery, right? You know, like that's, you know, super important, right?

[47:17] **Audrow Nash:** Yeah. Don't want to hit that artery.

[47:19] **Axel Krieger:** And yeah, yeah, and then, you know, just help what should be the next step, especially in a bad situation, yeah.

[47:27] **Audrow Nash:** Cool. Okay. So that all sounds really really cool. From all the data that you've been given with all this, how, tell me some of the challenges with that because I imagine you're getting a ton of different procedures. It's like I imagine while it's wonderful to have so much data it's also probably a challenge to make sense of that data and use it productively because there's probably so much variation in that. Tell me about it.

### [47:30] Mixing training data

[47:55] **Axel Krieger:** Yeah, you're so Right, you know, like, I mean, that, you know, like, you know, it sounds awesome to get like this, you know, totally diverse, you know, robotic data, some more endoscopic, some more, you know, you know, surgical robots, some more, you know, just the camera following like all different, you know, robotic architectures, but then if you all try to combine it, you know, sometimes just, you know, decrease performance rather than, you know, increase adrenalization capabilities.

[48:20] **Audrow Nash:** Chaos.

[48:27] **Axel Krieger:** So, what we found often is that we have to like, you know, use for pre-training some of the more

[48:34] **Audrow Nash:** Specific. Mm hmm.

[48:34] **Axel Krieger:** Aligned, you know, robotic systems. So, you know, like a subset of the 50 institutions has been like, you know, the better kind of training recipe for specific tasks on some robots, you know. So, yeah, absolutely. That's kind of a big, you know, big task is right, you know, finding the right, you know, the right mix of training data, and then, you know, really finding the right episodes, right, the really good episodes to train from. So, that's a lot of the, you know, secret sauce going into the beginning to have a good, you know, policy.

[49:16] **Audrow Nash:** How, can you talk a bit about how you do it for that? I know it's a secret sauce but

[49:20] **Axel Krieger:** No, no, I, no, I mean, you know, secret sauce is probably the wrong expression, the challenge, I would say, yeah.

[49:25] **Audrow Nash:** It's the challenge. Yeah yeah. I see.

[49:28] **Axel Krieger:** So, you know, so, for example, right, if we want to perform a task with the DaVinci Research Kit, then, you know, it really showed that if we kind of train it on the super mixed with all different robot architectures, it did not, you know, improve the performance, right. So, but then if you pick, you know, very similar robotic architectures,

[49:48] **Audrow Nash:** I would expect. Yeah.

[49:53] **Axel Krieger:** Then it's beneficial to, you know, pre-train this on the variety of tasks, and then fine-tuning it on the specific task versus just learning, you know, on the specific task. And that's been kind of like, you know, it makes sense, right? I mean, it and it the data really bore this out as well.

[50:14] **Audrow Nash:** Huh. Do you think that We're gonna have like it seems like with language models like large language models like similar approaches seem to have been done for a long time in the computational linguistics or whatever it is. But then when you have internet scale data then you get pretty crazy performance seems like you can reason now it's an awesome tool for a lot of things. Do you think that we're missing that kind of like super large scale data do you think that would just kind of solve everything for you or is it like getting and I know that the challenge becomes how to get access to everything but do you think it's largely a data problem?

[51:00] **Axel Krieger:** I think that's, you know, a huge part of it, why it's hard to kind of replicate those generalis generalization, you know, results of LAM for sure. You know, it's a different magnitude of data, right?

[51:02] **Audrow Nash:** The case. Oh yeah completely different. huh.

[51:17] **Axel Krieger:** So, absolutely, that's, and there I think we just need to be, you know, smarter and, you know, smart and do this, you know, and get successful at a much, much lower regime and engineer, you know, solutions to still get, you know, really good performance.

[51:34] **Audrow Nash:** Yes. Yeah that's the way because there's not enough data so you have to like do feature crafting or break things down manually or whatever it might be so you can get the performance you want with the limited data that you have. What other procedures so gallbladder surgery step two very clear are you doing what other procedures are you guys working to solve for is it the rest of the gallbladder surgery or what kind of things are you aiming at as a company?

### [51:44] From dead tissue to live animals

[51:45] **Axel Krieger:** Exactly. Yeah. Yeah. So, yeah. Yeah. Over the last year, really, you know, we've been pushing to get from ex vivo, so, you know, just, you know, piece of the liver and the god letter to in vivo, right? So, that's a huge step for us. And so, we, you know, performed the first, you know, in vivo study. Really looks very good, trained on ex vivo, performing in vivo, generalizing across that with breathing. So, that is super exciting and, you know, and, you know, amazing. And then, you know, we are working to include the other phases of the god letter surgery, so the,

[52:35] **Audrow Nash:** Wow. That is.

[52:51] **Axel Krieger:** You know, dissection, you know, phases, so that's thick. And then, expanding to, you know, different robotic systems and architecture. Super exciting for us is kind of endo-luminal, endo-vascular, so endo means inside, vascular is, you know, blood vessels, so, you know, inside blood vessels. So, they're really nice new robots that drive, you know, catheters into various parts of your body, especially, you know, in your brain for stroke

[53:25] **Audrow Nash:** Really?

[53:27] **Axel Krieger:** Repair and aneurysm repair. And,

[53:29] **Audrow Nash:** That's crazy. I didn't know that existed.

[53:31] **Axel Krieger:** You know, there's, you know, there's a lot of need for, you know, guidance for visualization for, you know, autonomy to, you know, let's say you have a stroke somewhere remotely, how can you know, do this, you know, without access to, you know, the high-level stroke center anywhere nearby. So, a lot of really cool need around this.

[53:58] **Audrow Nash:** Huh. Man what a thing. So it's trying to like democratize healthcare say or surgery access when it's needed for this kind of thing.

[54:00] **Axel Krieger:** Yeah. Yeah.

[54:09] **Audrow Nash:** And that's so cool. How Well so that sounds like a very I guess what do you think the timing for that kind of thing like it sounds to me like it may be far out where we are automating more things and making like I mean the I don't know much about the steering and the scope to block out stroke and stuff like that's so cool I've seen some where they do it in the heart. But in the brain is like bonkers that can be done I don't know too much but that seems crazy to me. What do you imagine for the timing for these kinds of things and just like this area in general where do you imagine it going?

### [54:47] Self-driving analogy and FDA approval

[54:52] **Axel Krieger:** Yeah. So, you know, I think a good analogy is there, you know, our journey in the autonomous driving world, right? We didn't go from, you know, zero, you know, autonomy zero, you know, navigation help to, you know, self-driving, fully self-driving car without a steering wheel, right? Like, we had like these little small steps and functions, right? And so, like a park assist and then on the highway a bit of, you know, self-driving, you know, a brake assist. And so, these are kind of like also the same kind of functions that are going to come into these surgical robots.

[55:08] **Audrow Nash:** Yep. Yeah. Mm hmm.

[55:28] **Axel Krieger:** So, and, you know, we already have FDA approved a autonomous camera following so that, you know, moon surgical is,

[55:36] **Audrow Nash:** Ooh.

[55:39] **Axel Krieger:** You know, the, you know, surgical device manufacturer that has FDA approval for their module that, you know, autonomously adjusts the camera position. So, this is a low-risk procedure, right? Because you don't interact with the tool directly and you can kind of limit the amount of motions to make it nice and safe. And so, you know, that's kind of how the FDA approves devices is by risk, right? So, you start with really low-risk, you know, functions and then, you know, we get more confidence, we get better, you know, and more robust and then, you know, can climb kind of that risk profile to, you know, more and more, longer procedures. So,

[56:05] **Audrow Nash:** Yes absolutely.

[56:23] **Axel Krieger:** Yeah, adoption of guidance, decision-making is going to, you know, tracking that's, you know, already kind of coming and, in you know, in some way, you know, already, you know, coming and autonomous action limited in, you know, functions like, you know, camera control and now some, you assistive, you know, traction or, you know, like some suction, you know, maybe then some suturing, maybe some resection and then eventually, you know, full procedures, but there needs to be definitely a lot more, you know, work done to make it happen.

[56:51] **Audrow Nash:** Some suturing maybe and. Mm hmm. Regulation. Yeah. Gotcha. Tell me about so you mentioned you went from whatever it is where you have a chunk of meat that you're like a dead chunk from the pig to in vivo alive pig or a live person with the gallbladder surgery and breathing. Tell me a bit about that like I imagine breathing is interesting because everything is moving maybe they don't breathe that fast but because sedated but tell me a bit about the challenges there.

[57:33] **Axel Krieger:** Yeah, so we work with, you know, Children's National, you know, Richard Shaw and his team there at, you know, down in DC on, you know, performing, you know, disc golf ladder surgery in life peaks and so that's, you know, ongoing study, hasn't been published yet, you know, but, you know, the initial results look really promising. So super exciting that when we had a policy, you know, trained, you know, just on ex vivo tissue that it worked kind of, you know, out of the box per se, you know, in the first animal study with breathing and motion was, you know, just super exciting.

[57:56] **Audrow Nash:** Coming. Yeah. Mm hmm.

[58:17] **Axel Krieger:** You know, definitely not easy. It's a different robot. It has different configuration. It is, you know, of course, the, you know, bleeding and breathing is, you know, different than what we have trained on. So, you know, exciting, you know, challenges, but, you know, really looks, it looks very promising for the overall framework.

[58:31] **Audrow Nash:** Mm hmm. Super cool. Where do you guys think you're going as a company because so you said funding late last year I think you said to me when we talked you started in February the company. Maybe I'm wrong with that but around then like early beginning of this year. What do you think your next year looks like. Maybe next three years too because it's still pretty early. But it seems like you're doing very exciting work. Where we going.

### [58:35] Inner Logic's data platforms

[59:12] **Axel Krieger:** Yeah, I mean, it's been a whirlwind of a year and with, you know, raising the funds and then, you know, really getting started and, you know, building our first, you know, product and, you know, and then, you know, working with different, you know, manufacturers. So, you know, it's, we don't have like,

[59:27] **Audrow Nash:** Have you launched your product. It.

[59:33] **Axel Krieger:** You know, we work with different, you

[59:34] **Audrow Nash:** The beta testing it or alpha testing it kind of thing.

[59:36] **Axel Krieger:** Know, we sign deals with, you know, manufacturers and we, you know, are deploying, but it's, you know, not like a, you know, strict, you know, product line. It's really customized for, you know, different device manufacturers, you know. So, it's, you know, a, you know, customizable, you know, a product line around,

[59:38] **Audrow Nash:** Yep. Nice.

[59:58] **Axel Krieger:** You know, a lot of data. So, you know, synthetic data and then, you know, also the other, you know, is what we talked about the struggle with how do you operationally, operationalize, you know, real data. How do you know, structure it? How do you know, make it so you can, you know, train with? And so, we have, you know, these two, you know, you know, data platforms as, you know, initial products.

[1:00:28] **Audrow Nash:** So that makes sense. That's exciting that it sounds to me a little bit like consulting work but it's like while you fit specific customers what they need and then eventually you generalize and that becomes a more general product for this kind of thing is that kind of the leaning you think.

[1:00:47] **Axel Krieger:** Yeah, let me just explain this a little bit better. On the, you know, synthetic, you know, data set, what we have is a large cohort of, you know, CT scans of patients, so tens of thousands of patient data sets with all various anatomies. And so, what we can do, for example, is to generate, you know, a large cohort of, you know, data with a certain medical device, you know, from a company, right? So, of course, there's customization by, you know, using our, you know, cohort

[1:01:22] **Audrow Nash:** Oh I see.

[1:01:23] **Axel Krieger:** With a specific device data. But, of course, all the, you know, staging and platform is reusable, right? And so, we can, you know, have the platform set up, you know, and then we can, you know, you know customize it around the device needs for, you know, a specific, you know, company. So, if you want to work with this on a, you know, bone fracture, you know, versus on a, you know, neurovascular guidance, you know, device, of course, it's, you know, slightly different, but it's still the backbone of it is this large data set, and, you know, then we can visualize it on your specific needs and with your, you know, in mind. And similarly, on the real, you know, data intake, right? If you work on an endovascular robot, it looks slightly different than on a laparoscopic, you know, robot, but it's still, you know, like, you know, annotating the anatomy, annotating, you know, the tools, you know, annotating the different steps of surgery are also, you know, very related, right? So, we support a wide range of procedural devices. It becomes a little bit customized per, you know,

[1:01:45] **Audrow Nash:** Mm hmm. Cool. Yeah. Application

[1:02:41] **Axel Krieger:** Application and customer, but, you know, on the back end, on our side, it's definitely, you know, general as a platform.

### [1:02:42] Simulated patients from CT scans

[1:02:41] **Audrow Nash:** Customer and everything. It's fairly general. Cool. Tell me about the role of synthetic data for this like I guess what you're saying with the CT scans is that like you have real patient data where they did a CT scan that gives you like a CAT scan. Is that one of those things where they sit in like a donut and they go scan through it and that gets there like 3D body decomposition or composition.

[1:03:12] **Axel Krieger:** Yes, exactly. So, it's basically an X-ray machine, but it rotates around you and has a very, very high resolution. So, you get, like, submillimeter resolution of the full body, the full anatomy of tens of thousands of patients. So, imagine, like, you know, like knowing all the different anatomical variations that your device could encounter. So,

[1:03:25] **Audrow Nash:** Cool.

[1:03:34] **Axel Krieger:** You know, like, if you do test your device in a cadaver lab, you can do it maybe dozens of times, you know, but you cannot get the scales of, you know, of tens of thousands, right? And so, that's, and our simulation is, of course, not just, you know, a statistical atlas of something, but it's real patient anatomy and capturing the real variations that we would see. So, that's the cool thing that, you know, we can use this for, you know, you know training models and then, you know, a lot of the testing. So, that's, you know, the exciting part to see this across variations.

[1:03:45] **Audrow Nash:** Simulation. Mm hmm. How is it used in all of this? Because okay so you have these really high fidelity three-dimensional models of people. How do you use those or where do you use them? Is it in simulating surgeries? Is it like I just the thing that I maybe naively think is the body is very deformable. And I think that would make simulation really hard but maybe you can just use it as a you kind of juggle or shift the organs around to where they are in this particular person and then you use that for variation and see if it will fail. How do you use simulation here? Like where is it really valuable in learning things for I don't know robotics or other applications?

[1:05:09] **Axel Krieger:** You're absolutely right. the you know depending on the application, it's either you know easy to use it directly or you need to do additional steps to make it usable, right? And So, you know, like, having a rigid, you know, body, you know, helps you already for kind of orthopedic devices, right? Your bones are not going to move and deform, right? So, you know, so we work, you know, with companies where, you know, like, you know, you have different, you know, orthopedic implants that you want to test on a large cohort of patients, right? So, that's, you know,

### [1:05:20] Simulating bones, vessels, and soft tissue

[1:05:33] **Audrow Nash:** Oh yeah you mentioned bones. Mm hmm. Oh that's so cool.

[1:05:49] **Axel Krieger:** So that's, you know, like, and then they get implanted on X, on the X-ray guidance. So, we can, you know, then, you know, do a physics-based, you know, simulation of a specific, you know, X-ray machine and can show you at these different angles how this would look like, you know, on a large cohort. We can vary the imaging parameters and then, you know, you know really help you in designing and accelerating your design and then, you know, building, you know, then models that help you track, you know, devices and, you know, so, and work towards, you know, more, you know, functions. So, yeah, that's easy on kind of bony anatomy. Then,

[1:06:35] **Audrow Nash:** Yeah harder on squishy anatomy.

[1:06:37] **Axel Krieger:** Yes, and some, you know, are also a little bit easier. We call this kind of like just deformation, right? If you don't cut, if you don't, like, you know, make larger changes. So, for example, endovascular, right, where you put like a catheter in, it like forms, but it's just, you know, a

[1:06:48] **Audrow Nash:** Yeah. Following something.

[1:06:57] **Axel Krieger:** Elastic deformation goes back to the original state, right? And so, it's also not that difficult to simulate. So, we can now simulate, for example, like a stand in a

[1:07:05] **Audrow Nash:** Huh.

[1:07:10] **Axel Krieger:** Vessel, right? We can have, you know, deploy, you know, different stands, you know, in these tens of thousands of, you know, models and it looks, you know, incredibly, you know, realistic,

[1:07:11] **Audrow Nash:** So cool.

[1:07:23] **Axel Krieger:** You know, in that model. So, we can, you know, functionalize it and then, you know, for the particular application, you know, increase the complexity of that model to really mimic the procedure. Of course, what we discussed earlier, the cholecystectomy, like cutting and stuff, that's kind of like the most difficult part, right? And so, there,

[1:07:43] **Audrow Nash:** Yeah I imagine.

[1:07:45] **Axel Krieger:** You know, we also work on really cool simulations starting with these, you know, large cohorts of data sets and new simulation techniques to model, you know, how to cut, how to suture, how to do this, and then how to render it, you know, that it's visually, you know, really similar to what we see under camera guidance. So, yeah, that's exciting work.

[1:07:51] **Audrow Nash:** Yep. Close. Is that? So what I'm imagining and I don't know if it's correct is this seems like kind of an adjacent value add to kind of the main quest of the company but it seems to me like eventually having access to all this data and having a lot of videos of surgery you can kind of put the two together. So I mean all of these body scans with a lot of surgery videos and you can use that maybe to get very accurate simulations or something like you could and then you would be less constrained by messy variable real world data for this kind of thing because you could procedurally try to generate a bunch of it or something. Is that kind of does it feel somewhat correct that there's kind of separate things that will eventually hopefully merge or?

[1:08:59] **Axel Krieger:** Yeah, absolutely. So, you know, we still need the real data to really capture the surgical intent and technique and skills, right? So, that's...

[1:09:08] **Audrow Nash:** Absolutely. And also how the body reacts and the deformable materials and when you cut what happens and breathing or whatever it might be.

[1:09:17] **Axel Krieger:** Yeah, yeah, but we are limited on how much we can, you know, collect there and we are limited on how many variations we're going to see. always going to be some crazy edge cases that we have to deal with, right? And so, that's where, you know, simulation, I think, comes really into place to,

[1:09:28] **Audrow Nash:** Oh yeah.

[1:09:34] **Axel Krieger:** You know, have, you know, a much larger cohort and include all the different variations that you have to, you know, deal with and encounter. So, that's where, you know, the two things do come together and really complement each other really, really well.

[1:09:54] **Audrow Nash:** How far out do you think that is the coming together? Like is it are there a ton of really hard unsolved problems there? Is it a data limitation? Is it all the things like how do you how do you think about the two coming together? Because it seems like that will be an amazing place to be in when that does occur. You'll just you'll be able to solve surgery at least for a lot of cases and then you can solve more from what you can solve because you're getting more data and they're being more widely used and it'll be a big flywheel. But when do you think that kind of thing occurs or is it very researchy and what are the problems before that?

[1:10:30] **Axel Krieger:** Yeah, and it really is going to depend on the complexity of the surgery, right? So, what we discussed earlier, the kind of, you know, bony surgery, we already showed that we can train something purely and it works on a cadaver, you know, like that, you know, segmentation of, you know, like...

[1:10:45] **Audrow Nash:** Do bone surgery. What are bone surgeries? Like what do you mean? Is it putting in an implant? Is it like cutting something off or yeah?

[1:10:54] **Axel Krieger:** Yeah, exactly, exactly. Or in evertebrae, if you need, like, you know, injections or, you know, you know that, yeah.

[1:11:02] **Audrow Nash:** Ee. You have to not be squeamish with all the things. With all the years in surgery robotics. Yeah yeah.

[1:11:09] **Axel Krieger:** Yeah, so, you know, the simulating the guidance, you know, for these, you know, procedures under X-ray work out of the box seem to real. So, that's, you know, already super, super cool.

[1:11:23] **Audrow Nash:** That's amazing.

[1:11:26] **Axel Krieger:** You know, then this, like, if you add some elastic deformation, what I described as, like, the stem deployment or a catheter

[1:11:33] **Audrow Nash:** Yep. Things where you're just viewing it.

[1:11:36] **Axel Krieger:** Going up, you know, into a vessel, so that is, you know, small, you know, a simulation. That is also, you know, feasible to simulate the physics a little bit harder. And we like to combine the real physics of real experiments with, you know, some that have more autonomous variations in synthetic world, but not as, you know, high fidelity on the, you know, physics interaction side. So,

[1:11:45] **Audrow Nash:** Simulate.

[1:12:05] **Axel Krieger:** Combining them together is really effective there. So, that's going to be the next one is these procedures like catheter guidance, like stem deployment with an elastic deformation. That is where the two, you know, going to come together next and where we still have to show, you know, this, but that's going to be the easier case than the, you know, real surgery cutting, you know, bleeding where the, you know, our current way of simulating it is still not quite there yet, you know, like, you know, it's

[1:12:27] **Audrow Nash:** Mm-hmm. Cutting and whatnot. Mm-hmm.

[1:12:41] **Axel Krieger:** Progressing, we're working it and the company is as well. So, you know, it's really, you know, I think a matter of time before it becomes, you know, strong enough to really, you know, mimic it. A lot of cool work with the world and on the world model side, especially I find and on the material point methods also the kind of two building blocks that you're really bullish about.

[1:13:03] **Audrow Nash:** Hell yeah. Really cool. Let's see. Going a bit bigger picture. How do you imagine surgery robotics proceeding? Like what you guys are doing? Any other interesting companies or research that's going on? Like how do you imagine it progressing like three years, five years, ten years from now? I don't know if we can even guess really but it seems to me like everything seems to be speeding up in terms of progress. I think these language models and just like intelligence being kind of on tap is making it so advancements seem faster. So I might make it extra hard to predict but what do you think? Like or even like chronologically how the things will fall. Like how do you imagine looking out in this general field?

### [1:13:07] Where surgical robotics is heading

[1:13:58] **Axel Krieger:** Y yeah, I think it's gonna really transform. I think it's gonna accelerate, I think it's going to come into the clinic from really, really low risk to high risk. I think a lot of data analytics and then guidance, then decision support, and then action. I think you know that it's already pouring in. you know. So you know, so yeah, If I see where the big you know innovations over the next 10-20 years, I think it's going to be a bit less on the new mechanical improvements. I think we are a little bit saturating there and more on the intelligence autonomy side. So, you know, I What we're going to see in three, five years are

[1:14:44] **Audrow Nash:** Probably.

[1:14:57] **Axel Krieger:** Some you know you know proof that these autonomous subfunctions are helpful in surgery. Imagine if you have you know, a one of the robotic arms you know s doing the blood suction to clear the field autonomously. So that th Those are functions that are technically feasible in a short amount of time and super helpful to surgeons.

[1:15:19] **Audrow Nash:** Mm-hmm. So cool. Mm-hmm. Yeah because if you can make it so they don't have to pay attention to all the things and it's like they have a trusty assistant that just does it and doesn't get distracted. Do you think autonomous cars are kind of the good metaphor for this? You brought it up. It does seem like it'll be a slow creep of or not maybe not slow but it'll be a creep of features until the whole thing is largely automated. And then maybe surgeons will be supervisors for this kind of thing.

[1:15:55] **Axel Krieger:** I think that analogy is really good. We are not going to be you know, fully manual and then suddenly you know, the you know, three years later it's like you don't need to search anymore. That's just not...

[1:16:01] **Audrow Nash:** Won't be pow autonomously. Probably still want the surgeon but

[1:16:12] **Axel Krieger:** Absolutely.

[1:16:13] **Audrow Nash:** It's like a power failure or the thing needs to be restarted or something stupid.

[1:16:16] **Axel Krieger:** Yes, no, I mean, we really, there's no desire to replace surgeons that's you know really making surgeons more efficient, effective, and performing this. We're just really

[1:16:28] **Audrow Nash:** Mm-hmm. I mean maybe.

[1:16:33] **Axel Krieger:** Helping with this crisis. We have fewer and fewer surgeons available. We have an aging society. We have a gap of in the next 10 years we are missing 20,000 surgeons, so we need to make them more efficient.

[1:16:46] **Audrow Nash:** How many surgeons are there?

[1:16:49] **Axel Krieger:** I would have to look it up, but I

[1:16:52] **Audrow Nash:** I'm just wondering if it's like if it's 10% more than we have or 20% or 30%. It's

[1:16:53] **Axel Krieger:** Don't know it. It's a big amount, but I don't know off the top of my hand.

[1:16:59] **Audrow Nash:** A large amount. I mean you see the labor shortages everywhere. Yeah, yeah. Uh-huh. What a time. The I wonder. It's good. I think it maybe will parallel programming a little bit too in terms of like I mean like I'm a programmer. I am writing a lot of code with agents now which is really cool when I supervise and I don't like that and I adjust things and whatnot. And it's like maybe eventually. Do you think that surgeons will kind of just like they'll watch the operation and they'll get results. Maybe they'll even do like a few surgeries at once because the robots and AIs are doing a good bit of it and they check in and make sure things look good. Do you imagine that would be kind of similar or what do you think?

### [1:17:28] One surgeon, several surgeries

[1:17:48] **Axel Krieger:** Yeah, I had at our surgical robotics conference this summer SRS a nice debate with Intuitive, with you know John Walsh there. It was interesting where we debated how a supervision could look like in the future, right? And you know and so My vision was eventually you might have a surgeon supervising several procedures at the same time.

[1:18:15] **Audrow Nash:** Yeah, yeah like a grocery store checkout model for surgeons. What a world.

[1:18:18] **Axel Krieger:** I think that's kind of like if you're incredibly successful and this it really, you know, is as good as it could be, that's kind of like the ultimate limit. Where you have a surgeon kind of like you know sitting on a couch and watching several procedures going on at the same time.

[1:18:29] **Audrow Nash:** Yeah. And like talking to the AI be like oh on screen two you need to be doing this or that looks a little worrying to me in this one. Make sure we address it. Whatever.

[1:18:45] **Axel Krieger:** But it's still super far away and right now I think you know like really the small little functions where you have direct supervision and surgeons can take over. That's kind of like the vision for the next few years.

### [1:18:57] Training the next surgeons

[1:19:03] **Audrow Nash:** Mm-hmm. I have a friend who's a surgeon and what he was telling me is that a lot of the older surgeons do not like the robot. Like I think they use them but they're not as proficient with the robots whereas he is like super good with the robots for actually doing surgery. Really likes it. This kind of thing. How do you imagine this changing over time? Like I'm thinking with a lot of AI it's like technology is actually getting more accessible. Do you think that this will help to narrow this gap where it's like newer surgeons embrace it real easily and kind of older ones may I don't know be hesitant to go down this robotic path or what do you think?

[1:19:48] **Axel Krieger:** Yeah, y yeah, I mean, you know, surgical robots are already in a lot of the you know disciplines are already kind of part of the training program, right? If you look at the urology trainings, the residents kind of like, you know, use it very early on in the training. So I think of course the you know that that's changing how we

[1:19:55] **Audrow Nash:** They're everywhere yeah.

[1:20:14] **Axel Krieger:** Train you know surgeons. So yeah, I think, yeah, there's kind of like co-role in helping, you know, and learning from the beginning how it can improve and what are the pitfalls. I think it's going to be also important to include that in the training program. You know, of course, you know, becoming a surgeon is super difficult, you know, like years and also,

[1:20:41] **Audrow Nash:** Takes years. Yeah. Ten years or something more.

[1:20:45] **Axel Krieger:** You know, it also is a combination, right, between, you know, all the cognitive skills, but then also all the kind of fine manipulation skills, right? And so maybe if we, you know, capture some of those fine manipulation skills with, you know, with more robotic technology, maybe it opens up the profession for also, you know,

[1:21:07] **Audrow Nash:** More folks.

[1:21:07] **Axel Krieger:** Other talents, you know. So yeah, maybe there's a way of, you know, long term shift, but I, you know, haven't really deeply thought about it or...

[1:21:09] **Audrow Nash:** Hell yeah. Seems like it could be. I mean especially like I don't know I wonder what do you think will be the way that features will fall in the space? Like and then because I imagine all these will be very valuable and I mean like when will we get autonomous suturing? When will we get autonomous like simple procedures like an appendectomy like cutting off the appendix or something like that or when will like these kinds of things? What are a few milestones and like I'm sure it's very hard to predict the years and everything but like I don't know ballpark of how you see them in terms of difficulty and maybe timeline.

[1:22:04] **Axel Krieger:** Yeah, I think a lot of that, you know, depends, of course, you know, on kind of the funding and who's going to like, you know, really, you know, take this on and with, you know, how much of an effort a specific task. So it's like hard for me to...

[1:22:15] **Audrow Nash:** Wait. Mm-hmm. How is the funding climate for this would you think like you mentioned investors reaching out to you they're looking for other things which is really awesome. How and for you all it was like a pre-seed funding or like I know it's a medical related one so probably a much larger pre-seed than most. Was it pre-seed or seed round?

### [1:22:19] Funding climate

[1:22:39] **Axel Krieger:** A seed round, seed round. Yep.

[1:22:41] **Audrow Nash:** Okay but for like Series B appetite and like things where they get the companies become much and much more established. How is the general appetite for investing in this like medical technology? It's probably growing I would imagine but what do you think? What's your read on the environment?

[1:23:05] **Axel Krieger:** Oh, I mean, I think it's, you know, a fantastic environment growing. I mentioned that there are more surgical robotics companies coming in. So it's a highly dynamic, you know, environment. You know, this kind of surgical robotics conference, the SRS conference, has been like seemingly doubling in size every year. It's kind of crazy how it's expanding. I mean, it's really massive,

[1:23:14] **Audrow Nash:** Yeah. So cool.

[1:23:32] **Axel Krieger:** You know. So there's a lot of, you know, appetite for an excitement around, you know, AI and surgery. So, you know, I think that's definitely...

[1:23:41] **Audrow Nash:** Mm-hmm. And big aging population need a lot of medical treatments are at the end of people's lives which big aging population so there's gonna be a lot of diet or a lot of appetite for lots and lots of surgeries probably. And big aging population means probably lots of investment in medical stuff. It's probably it's probably a very good time for surgical robotics. Would you think?

[1:24:09] **Axel Krieger:** Yeah, wonderful time. Yeah. And I think, you know, our approach is not to become, you know, the 56, you know, vertically integrated robotic system, you know, but, you know, horizontally support, you know, the AI roadmaps of, you know, you know, s all different, you know, robotics surgical companies. So for us, because we are in this supporting role, I cannot give a prediction how long a certain function will take because, you know, that really is driven by, you know, the, you know, you know a certain robotic company like I'm gonna, you know, really focus on this, you know.

[1:24:45] **Audrow Nash:** Do this one but yeah.

[1:24:47] **Axel Krieger:** That's why it's hard for me to predict with a really exact time scale, you know. Yep.

[1:24:50] **Audrow Nash:** With any certainty. Yeah cuz you don't have control of it. Do you have any do you have any guests though like what all the milestones be over the next few years? Doesn't need a timeline but what do you think?

### [1:24:52] Next milestones, from suction to suturing

[1:24:54] **Axel Krieger:** Yep. Yeah. Yeah. What I mentioned is that, you know, autonomous camera control is already FDA approved, right? And so I really feel like in the next year or two, I would be, you know, really shocked if they're not more of these smaller low risk assistive functions coming in. So what are some Targets are, you know, like some, yeah, yeah, exactly soaking up blood. And then, you know, like the kind of grasping, you know, like small changes of the grasper can also be kind of limited in the amount. So like the active grasping. So it kind of like exposes, you know, different anatomy for you to exactly, you know, like holding the liver up, something like that in a, in a gentle way, like for a certain procedure. So that is, I think, you know, coming. I I'd be shocked if that's not coming in the next couple of years, you know, but, you know, after that, you know, full, you know, suturing anastomosis, that's a little bit, you know, exciting.

[1:25:06] **Audrow Nash:** Are you guys cool? Soaking up blood or whatever. Yep. Got like peek under something? Yeah. That would be exciting. What's the esteromosis? What does that mean?

[1:26:01] **Axel Krieger:** And as the most is the closing of tubular structures so that, you know, like, yeah, yeah, exactly suturing off, you know, tubular structures together, where you have to do a like suture all the way around. If you know, have a little gap, that's where you have a leak. Yeah. Yeah.

[1:26:05] **Audrow Nash:** Oh like intestines or something. No good. Gotcha. Yeah suturing. I feel like I don't know anything but I feel like suturing will be a big win when we get there because it's so repetitive and I do feel like quite a lot of errors happen in it

[1:26:28] **Axel Krieger:** Hmm. Absolutely.

[1:26:32] **Audrow Nash:** Overall. Just it's so tedious and you can have so many required. So I don't know. That would be great to automate.

[1:26:39] **Axel Krieger:** Yeah. Absolutely. I think that's a, you know, a very good goal and there, you know, there are some procedures that are a little bit lower risk like closing off the port after you take out that little port, you have to close it up so that's, you know, a low risk, you know, suturing procedure, where would be a really good starting point, you know, yeah. Yeah.

[1:26:53] **Audrow Nash:** Yes. Yes. Hell yeah. Awesome. What do you think is after suturing? Like what gets a little harder than that?

### [1:26:58] Tumor resection and full procedures

[1:27:08] **Axel Krieger:** I, you know, and Johns Hopkins we work a lot in resections, so, you know, tumor resections, you know, like, that kind of, you know, precision tumor resection you know where you pay attention to not, you know, hit any, you know,

[1:27:17] **Audrow Nash:** Oh so removing something. Arteries or veins or whatever.

[1:27:27] **Axel Krieger:** Vessel or artery or, you know, any, you know, nerve, and, you know, need to fully remove the tumor. You know, it's so hard for surgeons right you have like a preoperative scan like an MRI or CT, and you mentally kind of map this together in your head together what you see during surgery. And so I really think a robot that, you know, does this registration really well learns how to do these cuts can, you know, potentially, you know, do this, you know,

[1:27:51] **Audrow Nash:** Feels very well suited for it. Way better, faster. Mmhmm.

[1:28:01] **Axel Krieger:** You know, a bit, a bit better. So I think there's definitely potential.

[1:28:04] **Audrow Nash:** Because you can have it map directly the scan over what it sees which that'll be like a perfect mapping. I don't know. I just you hear things about like the wrong leg being amputated or the wrong like things like this. So it's like I know surgeon like those are freak cases but they are busy. They work long shifts. Surgeries can take a long time. I don't know a robot that just transposes it and is like oh we're gonna cut this out like that could be really really good. Hell yeah. And then what's harder than that? Maybe we'll stop at this one but what's the what do you think is the next thing after removing chunks? I don't know how to phrase things right.

[1:28:49] **Axel Krieger:** I think, you know, you know, removing the tumor is still like one step out of a full procedure right because if you do the procedure right you have to like plan where you go in, then you have to like kind of dissect to get to the anatomy of the tumor. You have to like, you know, remove all the fat, you know, create that exposure, you know, then, you know, maybe, you know, you know, stabilize it so like it just becomes this kind of long,

[1:28:57] **Audrow Nash:** Yep. Mmhmm.

[1:29:18] **Axel Krieger:** You know, thing that all you know needs to kind of fit together, you know, beautifully. So that's, I think that, you know, that the most difficult endpoint, the peak point would be to do like full procedures, you know.

[1:29:21] **Audrow Nash:** Mmhmm. The peak thing. That would be so cool. Imagine like you just go somewhere and you get an operation you needed this kind of thing. Like buy an autonomous system that'd be so cool. I wonder if that'll happen. Do you think that'll happen in like... Would you be surprised, I guess it's kind of a ridiculous question, but would you be surprised if it wasn't the case in 20 years from now? That you could just like go and have a robot operate on you? I mean the like market side of it to be determined but like fully automated surgeries in 20 years. Would you be surprised if it wasn't the case? It's hard to predict without technology is now but...

### [1:29:49] Robot surgery in 20 years?

[1:30:05] **Axel Krieger:** Yeah, yeah, I mean, right now with where, you know, looking at the trajectory and being in it, I would be surprised if there are not at least some procedures where that's the case, you know, like what we discussed before there's like a nice,

[1:30:15] **Audrow Nash:** Better.

[1:30:19] **Axel Krieger:** You know, like

[1:30:20] **Audrow Nash:** There's some easier ones. Mmhmm.

[1:30:21] **Axel Krieger:** Scale of complexity, you know. So some of the endovascular procedures, you know, might be the first ones and, you know, even with the kind of surgery like cholecystectomy, there's also different scales of complexity. Some are elective where you don't have much inflammation. And so those are easier to perform than the ones that are, you know, where your gallbladder is all angry and all red and swollen up. Those are super hard, you know, so, you know, there's, I think these, you know, simpler, you know, ones, I can certainly see in that timeframe.

[1:30:29] **Audrow Nash:** Mmhmm. Yeah. There's levels. Mmhmm. How would you... So you're obviously in an amazing position in this really exciting time for robotic surgery. How would you... If you were someone say you're like done with college or finishing college or something like this or early in your career and you wanted to go in the direction of surgery robots or surgical robots, how... What advice do you have for them I guess? Like how would you start to get on this? Because it seems like such an exciting time.

### [1:31:15] Advice for getting into surgical robotics

[1:31:29] **Axel Krieger:** Yeah, I mean, I can just speak from experience that this is really my passion, right? I feel in robotics, right? You know, working in healthcare and helping patients is just such a, you know, really cool field to be in, right? So if you're kind of like thinking, I love robots and, you know, which field is for You know, I'm so incredibly happy to have landed, you know, in the medical, surgical, robotic space. It's just, you know, so rewarding. You know, through my PhD, I built a robot, you know, and we, you know, was used on a patient and I talked to that patient afterwards, right? And that was kind of a transforming moment for me, right? Like kind of like, you know, so that high of that, you know, feeling that is something where, you know, it's just not, you know, in other fields, you know, the case.

[1:32:11] **Audrow Nash:** Whoa. Oh I imagine. Ha! I mean I like when people use my software. That's completely different. Like you've built a robot and operated on someone. Like that's bonkers.

[1:32:32] **Axel Krieger:** Yes. Yeah, yeah, that was, you know, incredible. So...

[1:32:40] **Audrow Nash:** I bet you're one of the few people in the whole world that can say something like that. That's crazy.

[1:32:45] **Axel Krieger:** Yeah, really, really lucky. So I, yeah, don't take it for granted. And I just absolutely, you know, love to share the passion and why I think this is so exciting. So, yeah, I think it's super important to find your passion, right? And then, you know, if you want to get into surgical robotics, right? Yeah, it's great because there are several entry ways, right? More clinically, more, you know, hardware oriented, more from a, you know, ML, AI side, right? There are so different entry points into it, right? And so, you know, that's, I think, super exciting. And you have to have a bit of this kind of system mindset, you know, for all the different things to work together.

[1:33:29] **Audrow Nash:** Hmm. Any advice on growing that system mindset? Because that seems like a hard thing. Like just mentoring students and stuff. It's like that's a really hard thing to teach in my experience.

[1:33:39] **Axel Krieger:** Yeah, you know, I think, you know, we, our company's right next to Hopkins Hospital and, you know, with, of course, you know, this great relationship, we can see procedures, right? So, you know, I think that's, you know, just so amazing. If you know, can, you know, really see that how this is being used, right? And, you know, now with, you know, YouTube videos, right? You can also get a really, really good feeling for the different procedures. You can get a lot of, you know, access to, you know, how things are performed. So, yeah, it's, you know, dive deep into it, you know, on the use case, you know, understand, you know, that the language, right? So you can talk with researchers, understand video, right? So that's, I think, you know, a really important aspect of it. But you can find your way into this, you know, more on, you know, from a clinical perspective, more from a, you know, device, you know, robot architecture perspective, or, you know, more from a AI, ML, you know, training perspective. But it needs a bit of the three in combined, in combination to be, you know, really e effective.

[1:34:52] **Audrow Nash:** Mmhmm. Yes. Yeah you can't forget one. It just doesn't work. A total non-starter. It's like a stool fallover or something. huh. Let's see. So if someone watching this takes away one thing from our whole conversation, what do you want it to be?

### [1:35:02] Final takeaway

[1:35:26] **Axel Krieger:** So, I it's something think less on the, you know, s robotic architecture side. It's really the next, you know, decade will be on, you know, intelligence autonomy. And so,

[1:35:30] **Audrow Nash:** Hardware side? Mmhmm.

[1:35:39] **Axel Krieger:** Yeah, that's the, you know, direction and the change, the transformative change that is happening. I think it's, you know, fantastic for, you know, access for patients. And, you know, super excited to be, you know, a part of this change.

[1:35:58] **Audrow Nash:** Oh yeah. Alright thank you. It's been a lot of fun.

[1:36:01] **Axel Krieger:** Thank you so much. Thank you so much for having me. It was awesome.

[1:36:04] **Audrow Nash:** Oh yeah. Alright bye everyone.
