Building a More Connected Healthcare Workforce with Amit Kumthekar, Co-Founder and Chief Product Officer at Sully

[00:00:03] [SPEAKER_00] Hello, everyone, and welcome back to the Outcomes Rocket podcast. So excited you tuned in to another episode. I am thrilled to be having a conversation with an incredible person and an innovator in healthcare. His name is Amit Kumthekar, and he is the Co-Founder and Chief Product Officer at Sully.

[00:00:28] [SPEAKER_00] They are doing extraordinary work in the AI and agentic space, removing a lot of fluff from workflows, making it easier for clinicians to practice. The vision is one doctor, one patient. I am excited to have him here to tell us all about it. Amit, thanks for joining.

[00:00:46] [SPEAKER_01] Great. Thanks for having me. Super excited to chat and share a little bit about what we are doing.

[00:00:51] [SPEAKER_00] Yeah. Well, excited to learn about it, Amit. And before we dig in, tell me a little bit about yourself. What got you into healthcare and what keeps you in healthcare?

[00:01:00] [SPEAKER_01] Yeah, no, I think that's a great question. It's kind of where I tend to start. So I was, and then kind of where I started, I was diagnosed with type 1 diabetes when I was two years old. So later actually found out that I had something called monogenic diabetes due to a misdiagnosis, some mislab tests when I was young. But growing up kind of with a chronic condition, I realized I've experienced healthcare my whole life. I wanted to do something in the space. So when I was eight, my dad and I started jailbreaking insulin pumps.

[00:01:27] [SPEAKER_01] So at the time, we were doing things with forecasting models to predict blood glucose, using that to then tell these kind of jailbroken insulin pumps when to give insulin dosages. So that was an open source project actually in the Bay Area by a bunch of Pixar engineers. So we contributed to that, which actually spun out into a company. But that was kind of when I knew I wanted to do something in healthcare. That project didn't go on for long because my mom shut it down because it was unsafe. We were jailbreaking these insulin pumps and she was like, no.

[00:01:56] [SPEAKER_01] So biohacking, right? You're biohacking. You know, we were the OG biohackers. But yeah, that community was wonderful. It was kind of shed light on the ability to use technology to improve healthcare and improve the lives of individuals. And that's when I knew I wanted to do something in healthcare. So started doing more research in diabetes. I started in high school at a lab at Stanford doing monogenic diabetes research. Continued that through my undergraduate at Stanford, but felt like wet lab and computational research was too far detached from patient care.

[00:02:25] [SPEAKER_01] That work is incredibly impactful, but it's a very long horizon. So your work is going into papers, going into patents, which then are spinning into startups or going into pharma. And that horizon is 10, 15, 20 years. And for me, that impact is great, but I wanted to have more direct impact. So started doing clinical research, became one of the youngest, if not the youngest principal investigator at a company called Dexcom. And Dexcom is doing continuous glucose monitoring. So got a decent amount of funding to do CGM research in underserved communities.

[00:02:54] [SPEAKER_01] So I did that in the Bay Area, actually at a clinic that I'd volunteered at when I was a student at Stanford. And that was kind of when I realized I really want to do something with healthcare, using that to impact even more people, because that study was impacting hundreds. But I knew I wanted to do more. So funny enough, ran into Omar, my co-founder at a cafe in downtown Palo Alto.

[00:03:16] [SPEAKER_00] So you guys didn't know each other? You guys just ran into each other?

[00:03:19] [SPEAKER_01] No, I actually had read an article about him and I recognized his face. And so I was 18 at the time. And I remember telling my girlfriend, like, look, I know who this guy is. And he leaves the cafe. And frankly, I was nervous. But I remember telling my girlfriend, like, I'll never see this guy again. Like I might as well. Right. So he leaves the cafe and I run out of the cafe, run across the street, go to another cafe.

[00:03:42] [SPEAKER_00] And I'm in. I have to ask you something. So what did you know Omar for before you met him?

[00:03:46] [SPEAKER_01] I wrote an article about Odigo, which was his previous company. So yeah, funny enough, just a YC article that I'd write about them.

[00:03:52] [SPEAKER_00] You ran after him. And then?

[00:03:54] [SPEAKER_01] Yeah, I ran after him, entered this cafe, bugged him because I was very interested in research and saw I was interested in what he was doing. And so we chatted. And funny enough, we just hired one engineer, which he met in that same cafe. No way. And that guy's killing it too.

[00:04:11] [SPEAKER_00] That's awesome, man. That's awesome. There's something magical about that cafe.

[00:04:15] [SPEAKER_01] Exactly. We're joking that sometimes we have to go, well, we go there frequently, but we have to give the owner some like equity and sully or something.

[00:04:22] [SPEAKER_00] That's so funny.

[00:04:23] [SPEAKER_01] But yeah, crazy story and very excited to have started building this company. And yeah, we've done a lot of cool work and very excited for the future.

[00:04:29] [SPEAKER_00] I love that chance meeting that turns into something. And, you know, you sort of had this thing that we all have. It's this fear of rejection. And then we don't do it. But you know what? You said, I'm going to do it. I'm going to go say hi. And so here we are. What's the cafe name?

[00:04:45] [SPEAKER_01] Okay. It's actually, okay. Free sponsorship. No, it's called Oklava Cafe in University Ave in Palo Alto. So, okay. I mean, yeah. Oklava? Oklava. Yeah. Turkish Cafe.

[00:04:57] [SPEAKER_00] Shout out to Oklava Cafe for the magical connections that you guys are making. Not just brewing coffee, you're brewing magic in companies. So that's awesome. Thanks for sharing that. So fast forward to today, you guys are doing some incredible work. Talk to us about what the company does and how you're adding value to the healthcare ecosystem.

[00:05:21] [SPEAKER_01] No, definitely. So I think I can summarize what we're doing, what we're working on in one line. And it is, we're building products and we're building technology with the goal of increasing the human capacity of healthcare systems. So we are attacking this problem kind of from two main fronts. The documentation aspect, which involves things like clinical intelligence, billing, documentation, clinical notes, this kind of stuff, which has become a bit more commoditized, a bit more standard.

[00:05:48] [SPEAKER_01] But then we're also approaching it from the front office side of things. So doing inbound calls, doing outbound calls, doing all kinds of patient engagement. And our unique selling point is really being able to combine both of those two mediums, more or less, to deliver a solution which has fewer points of friction.

[00:06:07] [SPEAKER_01] In healthcare right now, regardless of what kind of system you go to, if it's an enterprise AMC or it's a kind of mom and pop clinic, they are overwhelmed by point solutions. And there is no kind of unifying technology, which allows you to connect what your clinicians are doing to what your patients are experiencing. And so the goal of Sully is to move towards a future where you have a more connected workforce, you have a more connected kind of set of products.

[00:06:34] [SPEAKER_01] And for us, that's a medium not only to improve the experience, so to care about how people interact with healthcare, because that's fundamentally what drives us. Not only improving how patients interact with healthcare, but improving how clinicians interact with healthcare. And then aligning it with the incentives of hospital administration, because that is the nature of this industry. And so doing that on the billing side, doing that on the efficiency side, right? Being able to allow administrators to do more for their communities, right?

[00:07:01] [SPEAKER_01] That's kind of our thesis and what we've really engraved into the product and into the team at Sully.

[00:07:07] [SPEAKER_00] I think that's fantastic. So you mentioned on the one hand, the documentation, the intake. On the other hand, the workflows, the outbound, the inbound. How is what you do different or better than what's already out there?

[00:07:22] [SPEAKER_01] I think the first thing is that it's connected. So Sully was built from the ground up as a central platform that allows you to run your full kind of clinical operations. Now, obviously, we're adding on pieces as we grow and mature as a company. But the ability to have, you know, a phone call that you have with your patients, outbound text messages, all this stuff translate to your clinical tools is something that we haven't seen on the market.

[00:07:49] [SPEAKER_01] And it's something that we feel will reduce the biggest points of friction in healthcare. So I can give an example, which I think articulates this well. Yeah. I mean, I was chatting with one of our folks that we just hired and he said, you know, before he joined that it was very small. But he gave an example of his wife who called and said, hey, like I had been to the doctor, seen this guy two weeks ago and been waiting on a discharge form. The current process is that we'll go to a receptionist, that receptionist will make a note. Now, that note looks wildly different across institutions.

[00:08:18] [SPEAKER_01] For some, it's literally a post-it. For some, it's somewhere in the EHR. For some, it's a spreadsheet that they can give to the doctor or the clinician. And those tend to get lost. So not only do patients not get what they need. Now, clinicians have to do more work when patients come back in and administrators can't capture revenue. So with Sully's tool, the idea is, okay, we're going to handle kind of your voice system, your patient engagement, your patient kind of front office aspect of things.

[00:08:47] [SPEAKER_01] So we will have full context on what the patient is going through. And because of that, we'll be able to more seamlessly integrate it into clinician workflows. And then down through that, more seamlessly integrate all the knowledge that we get from patient check-ins and all those kinds of things into the billing cycle. And so for that kind of full aspect, we're able to deliver a product which makes it easier to interact with healthcare, which I can talk about a little bit. Makes clinician's life easier. So clinician's life easier and more proactive. And it helps administrators justify a tool like this.

[00:09:16] [SPEAKER_01] So that's kind of our angle. That's our unique selling point. And at the core of all this is really just how much we care about experience and about delivering a product which feels incredible to use. And as we've matured as a company, I think that's been engraved even more.

[00:09:31] [SPEAKER_00] Yeah. And a lot of times when the technology is effective, it almost feels like invisible. Is there an element of that in what you guys are doing? Oh, okay.

[00:09:40] [SPEAKER_01] Absolutely. And I think this is a very interesting point that you bring. We've had a few things. So first, earlier today, actually, one of my team members who is leading our voice AI, our voice research, he sent a message where a customer had flagged a concern that they thought our AI was handling calls instead of a human. Instead, we found out that it was actually a human that answered it. Our patients were thinking that they were still chatting with AI. So they're chatting with a human thinking it was AI.

[00:10:08] [SPEAKER_00] Okay.

[00:10:09] [SPEAKER_01] Now, we've had this kind of aspect of delivering these tools that just fit into the workflow. So you cannot tell when you're engaging with voice agents or text agents, right? And that has made patients, obviously, from safety and confidentiality aspect, like made them feel good about using these tools, made it very easy to fit into their workflows. I mean, of course, we're transparent when people ask, right? Is this AI? All that kind of stuff. But that's one aspect of it being invisible. The second is going into hardware.

[00:10:39] [SPEAKER_01] So part of going into hardware, so HP, which is one of the biggest suppliers of hospital equipment in the world, they invested in Sully and we have a partnership with them to build hardware that is more deeply integrated in networks of clinics and also in the hospital room, but it's powered by Sully's intelligence. And so the idea there is that really becomes invisible, becomes part of the environment.

[00:11:02] [SPEAKER_00] That's great. What type of device? Anything in particular?

[00:11:05] [SPEAKER_01] I wish I could say more. It's a very exciting project, but it is very early stages. And I think we're going to definitely be chatting about it. It's a huge project with a bunch of major medical systems, probably one of the top two, three hospitals in the world are involved in it. And, you know, it's not my place to say it yet, but more to come.

[00:11:24] [SPEAKER_00] Okay, awesome. No, that's exciting for sure. And I love it, you know, and this merging of AI with hardware. You got Sandy, you know, software as a medical device. All these things are merging and you guys are at the forefront of it. So pretty cool to hear about this partnership. As you guys have been doing the work you're doing.

[00:11:47] [SPEAKER_00] I mean, I've seen some of the work you've done to emit around some of the multi LLM stack research and things like that. So none of this stuff is easy. You make a lot of mistakes. You learn more from those mistakes, I find, than the successes. So talk to us about biggest setbacks and key learnings from those. Or maybe one, right? One that comes to mind.

[00:12:08] [SPEAKER_01] I think we definitely have an interesting one at Sully. So like you mentioned, we do a lot of research. Clinical research, clinically applied research, AI research. And that's because a lot of us on the research team and applied AI team come from academic backgrounds. So I came from an academic background. A bunch of the other team comes from academic background. So we're a very research-driven org, which means we've been able to develop a lot of technology in this space, which is relatively ahead. Which means it's higher kind of accuracy. It is cheaper. It is faster.

[00:12:39] [SPEAKER_01] It's better to use. The problem, one of the biggest setbacks we made was investing a lot in the research, but less in the machinery, like the engineering aspect around it. And so what ended up happening is we developed these models, these harnesses, these agents, which were capable of doing pretty extraordinary things, like learning how clinicians write notes over time.

[00:12:59] [SPEAKER_01] But our ability to translate them into a product that was very robust, easy to use, relied on this engineering and machinery aspect, which we hadn't invested as much time into. So it slowed us down. Now we've learned from that, like, how do you do both at the same time? So we began to integrate researchers as PMs, working with engineering to be able to co-develop these things.

[00:13:21] [SPEAKER_01] We also spent more time kind of assembling a team which has strong engineering and strong research fundamentals as one person having both, so that they're able to take research to production much faster. But that was really a big learning because we've done all this cool work and we wanted us to get out, couldn't do it, right? Or it was going to take too much time. And to me, that was frustrating because it was a missed opportunity for our customers to get something sooner. But I think that is a lot of progress we've been making on that front.

[00:13:48] [SPEAKER_01] We've done, I think, frankly, incredible things on the voice experience side of things. Our research has been going to production with customers, with customers we have close partnerships with very, very quickly. And we've seen kind of drastic increases and drastic improvements on the efficiency. So that was a big learning and something that I think is building the character of the company, building the character of our or the culture on how we build products.

[00:14:13] [SPEAKER_00] Yeah, I love that. I think that's such a great example. And it sounds like now you've fixed that structurally where some of your research are sitting in PM roles and sort of both functional arts of the art are collaborating. And you mentioned voice. What's a nuance in voice that you feel like most people don't appreciate? And it could be from a product kind of leader standpoint, or it could be from a clinical leader standpoint.

[00:14:40] [SPEAKER_01] I think I'll give you one on the product side of things. And then I want on the clinical as well. I'll start with the clinical. I think one thing that when you think about this technology, one thing that humans are incredibly good at is adjusting the way that they talk to someone based on who that person is, right? Level of empathy. There's level of understanding on how to communicate with someone, right? Now, everyone has a different degree of ability to do that. Some people are more in tune on the social aspect of things.

[00:15:07] [SPEAKER_01] And what we found is that AI and things like that have not fully embodied what it means to converse like a human. So there's this nuance in talking to something that sounds like a human, responds like a human, but has something missing. It just feels off, right? I know exactly what you mean. Yeah. And part of that comes down to the flow as a whole of a conversation, right?

[00:15:33] [SPEAKER_01] How when you and I are speaking and I speed up because you're excited about something and I'll match that energy. When you slow down, I'll match that energy and vice versa. That we are working and a lot of research is going on to see how we can begin to emulate that. To be able to deliver experiences that feel better, that more empathetic, right? And we can consider how important this is on a clinical aspect. Because clinical aspect and clinical situations are when people are most vulnerable.

[00:16:00] [SPEAKER_01] When people are going through the hardest times in their life, right? Either for them or their family. And so it makes sense to give them an experience, regardless if it's human or technology, that will make them feel most at comfort, right? Because if they're most at comfort, they're going to have the best experience. Which means the healthcare organization serving them is going to have a better experience. It's going to be more efficient for everyone. So that's kind of a nuance there.

[00:16:24] [SPEAKER_01] On the technical side, this is an interesting problem that we've kind of encountered, which is how do you respond to something very quickly, like a human? So when you think about conversation, because we, funny enough, when we were building, like, and doing our research, we were just talking and saying, like, how do we talk, right? How do we have these conversations, right?

[00:16:45] [SPEAKER_01] One thing that was interesting to us was when you and I are having a conversation or having a conversation with anyone, generally speaking, I tend to form a thought on what I'm going to say next before you're fully done with saying something. So we wanted to begin to emulate that in our agents and in our harness. So how do we understand when to begin thinking, when to start and form something? And how do we either terminate that thought and switch to a new thought if new pieces of information come up or stick with it so that the return, right?

[00:17:12] [SPEAKER_01] So when a human ends and agent starts, how do we make that turn super fast? And so we've done very interesting work. Luckily, there's been huge advancements in speech to speech models recently. And so we've been able to build off of that. We've been working with NVIDIA on a few things. But the idea was to have a part of our machinery that was transcribing text also output when do we need to start thinking? When has a complete kind of thought or like a discrete point in this conversation closed? And when can we start kind of taking that and running with it for a little bit?

[00:17:41] [SPEAKER_01] So we're prepared to deliver a response after whoever's on the call ends their turn. So, I mean, those are two kind of nuances.

[00:17:48] [SPEAKER_00] Yeah, yeah, yeah. I love it. That I think are interesting, yeah. It's fascinating, right? And like what you said, it's like those missing things where you're like, okay, definitely AI. Something's missing. And it's these intuitive ways, sort of like the psychology of it and how we process it. What happens below the surface before it happens when it's set? And it's really interesting to hear how you and the team are thinking about that.

[00:18:14] [SPEAKER_00] I mean, I wanted to pick up on another thing that I've been hearing a lot more about that I heard you say that's worth visiting for folks out there that are thinking about these types of technologies, the harness. And as we've seen more agentic call onto the scene, you start to hear more about this harness. Talk to us about why that's important and just give us a little bit of nuance there too.

[00:18:41] [SPEAKER_01] Yeah, for sure. So I think harness really boils down to being the infrastructure that your model sits in. And if you want to take another analogy, your model, right, like GPT or Sully model, whatever, that is the brain. The brain on its own can't do anything. If you have no hands, you have no legs, can't do much. Your harness becomes that. So you take this brain, you put it in your harness, and now your model can do things. It can make tool calls, which means to operate certain functionality.

[00:19:10] [SPEAKER_01] It can retrieve information, which is other tool calls. So the importance of a harness is really building this skeleton, which extends the capabilities of whatever model you're going to put in it, right? It makes it be able to interact with the real world, makes it be able to do things that are useful. And it takes it from being a chatbot to being able to do something like humans. So a harness has been particularly important now due to like the rapid advancement in models as a whole. We're seeing that companies like OpenAI and Anthropic are releasing models all the time.

[00:19:41] [SPEAKER_01] But so is the open source community. And for us, so all of our models are based on top of open source models. It is super important for us to be able to deliver models' quality of our product at the pace of the frontier. And so developing a harness, which is going to be this kind of model agnostic framework that we can plug and play these models into, allows us to do that in a way that is cost effective and also very quick to bring to production. And it also just extends the capabilities of the model, right?

[00:20:08] [SPEAKER_01] So that's kind of the idea of a harness, what the importance is there. And it gives us kind of a central place to continue building our product on top of.

[00:20:18] [SPEAKER_00] That's great. I appreciate that. You guys are deep into this. But for a lot of people getting into it, they hear these words and we want to make these words reality. So thank you for simplifying it for us. This is fascinating. I've got to say, you guys are up to some really great things. If you think about the depth of technology and advancements you're working on, what's a trend or tech that you think is going to change healthcare as we know it?

[00:20:41] [SPEAKER_01] That's a good question. I think a trend that's going to change healthcare is there's going to be a lot of consolidation of tech tools. So one of the biggest problems in healthcare, in efficiency, either economic or purely time, comes from having so many disaggregated processes, either through tools, through people, through process, all this kind of stuff.

[00:21:06] [SPEAKER_01] And with the advancement of tools that are so intelligent and the ability to develop these tools very quickly in a very high quality manner, I think we're going to begin to see a lot of consolidation across kind of a few types of tools. So I'm not saying across a few types of companies, but I think there's going to be tools and stacks which all kinds of enterprises or companies across every part of the market are going to be going to consolidate to.

[00:21:32] [SPEAKER_01] And that's going to allow them to deliver their quality of care with the tools that they needed in a way that's more efficient and more cost effective. So I think that's going to be a trend primarily because now it's easier to build tools. It's easier to build higher quality tools, but it's going to be up to the people and up to the companies that can build experiences that are able to impact or deliver as much value across many parts of the spectrum in healthcare.

[00:21:58] [SPEAKER_01] So I really think that consolidation is going to happen quickly and it's going to be good for efficiency. It's going to be good for financial and economic efficiency.

[00:22:06] [SPEAKER_00] Yeah, no, I love it. I think it is moving fast and ultimately it is very fragmented right now. And I'm with you on that aggregation. So Amit, this is awesome. You know, and as we get to the end of this episode, gosh, there's so much more we could talk about. But what closing thought would you leave us with, you know, also help us understand what's the biggest common problem that people are coming to you for, for help. So let's combine those two as we wrap.

[00:22:34] [SPEAKER_01] Yeah, I really think the thought to leave with is these kinds of tools that companies like Sully and others are building, I think have always been like, how can we save healthcare money, right? But that's really not the idea. The way that we've approached it is how can healthcare find a way to reinvest more into its patients? And I can give you an example. So we've been working with a large amount of kind of, and this goes into the problem aspect.

[00:23:03] [SPEAKER_01] We've been working with a lot of customers who have offshore call centers and offshore teams. And they're very expensive. Frankly, it's a lot of people. They're very expensive. And that draws from the amount of money that they have to deliver care to the community. And these are either FQHCs, so they're kind of Medicare, Medi-Cal specific clinics, or they're kind of smaller clinics. And so we've been able to work with them to kind of replace some of this work, bring it more in-house at a fraction of the cost.

[00:23:32] [SPEAKER_01] Now, for them, all these executives we've been talking to at these companies are not just pocketing that money. They're taking that and reinvesting it back into their communities, reinvesting it into things like more hours open, more locations, things like that. Now, of course, there's a financial aspect to that where neither of us are naive. And these hospital systems are also there to make money. But it does so in a way that's aligned with improving healthcare, improving access to healthcare as a whole. So I think that's kind of a closing thought.

[00:24:00] [SPEAKER_01] And then the other aspect of that is the problem that people are coming for us to solve is an experience that helps clinicians and patients interact more efficiently. So we got that voice aspect, you have that documentation aspect, and they're sick and tired of going to five or six different vendors. And so coming to us who have a research-backed team, done a lot of work on the quality side of things, has been a relatively unique selling point. And so those are kind of my closing thoughts.

[00:24:26] [SPEAKER_01] And it makes me very excited about our customers, the future of healthcare, and being able to really use these tools to reinvest, to grow, to make technology and make healthcare more efficient in both major cities and major academic institutions and also all over the U.S., including in rural areas. So happy that I got to experience it firsthand and work with companies and clinics all over the place, including remote Alaska. So that's what's exciting to me. And that's kind of my closing thought.

[00:24:56] [SPEAKER_00] I love that, Amit. And there is so much opportunity and everything that you shared. I'm hearing access. I'm hearing equity. I'm hearing quality. And so for health systems and physician practices out there seeking a better way to scale and to do what you do more efficiently and effectively, this is definitely one to visit. What's the best way people can get in touch with you, Amit?

[00:25:26] [SPEAKER_01] For me personally, I think it's just my first name is sully.ai. And then our website is sully.ai. I respond to all emails. So if you're going to need anything, you can let me know. And yeah, it's an interesting time. It's a very exciting space. And so there's a lot of work to do.

[00:25:40] [SPEAKER_00] Outstanding. Amit, thanks so much for being with us. Just such an opportunity to learn about the work that Sully is doing to improve outcomes, improve efficiencies, improve care. Folks, in the show notes, you'll find all the ways to get in touch with Amit and the team at Sully. So make sure you check that out. And if something today resonated with you, make sure you share it with the friend. That's ultimately how we improve outcomes by spreading the word. Thank you all for tuning in. And Amit, thanks for being with us. Thanks for having me.

[00:26:10] [SPEAKER_00] Bye.