The Cost of Failure Has to Go to Zero (ft. Nish Khandwala)

If you were CEO, what would you do differently?

I would try to come up with some form of governance structure inside the hospital where you can rapidly fail with your AI, quickly catch it and change it. It is pointless to have a 5-year plan right now or even a 1-year plan in AI. You just have to be like, how can we be the most nimble and most agile?

That's one of the problems is that they will have this structure that's worked for them 20 years ago and they're trying to impose that same structure on modern Welcome to Digital Voices, where healthcare and life science leaders explore the real work behind transformation.

This podcast is about people, leadership, and the conversations that move healthcare forward. Now your host, Ed Marx.

Welcome to another edition of Digital Voices, and I'm proud to announce for the first time, We're actually doing visuals as well. So Digital Voices is moving from a podcast-only format to also posting on YouTube. Yeah, Nish, you're the first and I didn't even tell you ahead of time. I hope you're cool with that.

Oh, that's amazing. Well, I'm, I'm glad I was prepared for it, I guess, somehow.

We're, we're like number 4. It depends on the day of the week. So it, it, it could be as low as number 10, but number 4 in the world.

Oh, wow.

Technology for podcasts. And so my, my kids who are legit social media influencers, I mean, that's how they make their living. They're always pushing me like, dad, you need to be, you need to post this on YouTube as well.

You know, all of it, you know, yeah, get it on Instagram Reels and then some TikTok as well, clips from it. Well, hopefully I'll see a lot of that today. Let's see.

Yeah.

Yeah.

And we already on there, already on IG, on TikTok. Yeah. But now we'll have more visuals than just, uh, audio. So. I was really curious about your company and about what you do, and that's why we're connected. I know that you're doing some of this work, and that's how I know over at my old stomping grounds at University Hospitals in Cleveland, Ohio, as well as Cleveland Clinic, which is literally across the street. So it's a lot of fun. But Nish, before we get there, what songs are on your playlist? What kind of music do you like to listen to?

Oh, I'm pretty Pretty standard stuff. Like right now I've been putting Some Nights by Fun on repeat. It's the song that I used to get pumped up about the day. But then, you know, the usual mix of John Mayer, Taylor Swift fan here as well, and some Bollywood music.

Yeah. Tell me your origin story. Like where you mentioned Bollywood.

So I'm Indian. I grew up, I was born in India, but actually I did not grow up there. My family decided to move to the Middle East when I was 8, 6 years old. So I don't remember a lot of my time in India, unfortunately. But my dad's a chemical engineer, so he moved us to Saudi Arabia where the oil boom was happening when we moved there. Not a lot to do in Saudi. I used to, I remember, so we are also a nuclear family. It's just my parents and I, I don't have a sibling. So I remember just coming back from school at like 1 in the afternoon and then having absolutely nothing to do at all. Spent 8 years there. And then my parents realized that, oh, Nish is not a total idiot. And so they decided to take me to a place that had a few more opportunities and we moved to Abu Dhabi near Dubai. That's where they still live. And then I came to the US for my undergrad. I went to Stanford. I did my undergrad and graduate studies there, and that's where I fell in love with the problem that then became Bunker Hill, our company.

Yeah. So tell me about your languages then. So you pick, you speak Hindi first, but did you pick up Urdu or any?

Yeah, I actually did. So my mother tongue is Gujarati. It's a dialect or a language in the western part of India. I also know Hindi because when you're, right, when you're in India, you kind of learned that. I picked up Arabic, a little bit of it. I can write, but I can't understand it that well. And I gave up on trying to say Arabic because I couldn't just get the vocal cords to work the way I wanted them to. But I did learn Urdu for 8 years. Beautiful language. The script, it's so beautiful. And then obviously English. So yeah, I think I've missed out on some of the major languages like Like Mandarin and, and then Spanish.

Yeah. Yeah. Of course, AI is the most important language of all.

But you know, the way to work with AI is through English now. So I think everyone's kind of saved there.

Yeah, that's true. Before we jump into what you're doing today, are there any words that you live by, any quotes or mantras or, you know, things that sort of guide you?

I think I've been very fortunate in my life to have Found a cause or a problem worth getting out of bed for. And I think if you find that, what I've realized like is that if you find something like that, one, you're really lucky because it is really just like you kind of need to stumble your way into something like that. You can't like, you know, work your way into that. But once you do that, everything else seems to fall in place in life. You know, like you find a purpose. And so I think that, you know, if I had to like give any advice to my younger self, that's probably what I would say. It's like, find your purpose in life somehow.

Yeah, I love that. Was there a pivotal moment in life that fundamentally changed your trajectory?

Yeah, I, for the longest time, well, there are 2 things. One was when I moved from the Middle East to the US. Like imagine a kid who grew up in Saudi Arabia devoid of opportunities. Like there was nothing, nothing to do there. To someone who then has, the first place he landed in the US was Palo Alto, California. And then just like the sheer, I've seen both sides of the spectrum at this point. I am very grateful to be living in this country and the kind of opportunities that it allows, but also the people you surround yourself with. Like, I, it, you know, I was really grateful to the fact that the people who I was around always encouraged, you know, like we, you could have the craziest ideas and rather than try to put you like, oh, no way, that's not going to work. Or like, stay in your lane. It was always like, oh, that's kind of interesting. And then you sort of like, you know, work your way into that through curiosity. So that was life-changing for me. I think that was an environment, like with all the opportunities plus the encouragement and the support system, like that's something that definitely, like it's my life before the US and life after I moved to the US. That's one thing. But I think a lot of immigrants share that. You know, I think a lot of us would appreciate that. The pivotal moment that really like, oh, I can definitely see a life before and after that was my dad's heart attack. He's fine now, thankfully, but it really knocked the set of priorities into me. I think when you have a life event like that, it just sort of, nothing is ambiguous. It hits you directly and you have a rank list of priorities suddenly. And your value system is incredibly clear. Just the moment of clarity, very sad moment, but it just lends incredible clarity. And so that definitely, had it not been for my dad's heart attack, I don't think we would have embarked and continued on this journey that we've built Bunker Hill on.

Yeah. Yeah. So let's talk a little bit about that. I mean, you went straight from Stanford, right, into Bunker Hill.

That wasn't the plan. For the longest time, I think I wanted to be a PhD student who like did more research. I was a part of an applied AI group and in my preparation to potentially one day become a PhD student. And we used to build AI models for different use cases. In fact, one of the models that I last built in my role there was for a preventive cardiologist at Stanford. He came to us with an idea that we thought was like very cool. It was like, Oh, I see patients in my clinic usually after they've had their first heart attack, and my job as a preventive cardiologist is to prevent the second one. And so he told us, and this was shocking to us as people from outside of healthcare, that he would routinely see patients after they've had their heart attack and notice that the same patient had come to the same hospital 4 or 5 years ago, entirely unrelated reason. Maybe they had pneumonia and went on to see a pulmonologist. Maybe they had a car accident and went to the ER. They got a CT scan of the chest, and on that scan, it was very visible that there was a lot of plaque buildup in the coronary arteries. But unfortunately, the patient just fell through care gaps, you know? And so his idea to us was, hey, you AI guys, like, could you build us an AI model that could comb through every patient who is coming to the Stanford Hospital? see who's at high risk, who's at low risk, and for the ones who are at high risk, get them in front of a cardiologist ASAP as opposed to waiting for an event to happen. And so we built that model, you know, it worked really well as well. We published a paper in Nature and we boasted, bragged about how this is going to revolutionize cardiology, how this is going to save hundreds of thousands of lives. And I know you're smiling because I know where, you know, this is going. It went nowhere. It was collecting dust on my laptop 2 years later. That was my first experience of you build something that you really think has value, but it then goes nowhere. Of course, it was just embarrassing because, you know, we had made all these bold claims in the paper. But it was also very demoralizing because it's like, why work on something if you never really—

Yeah.

you know, had a plan to get it out into the real world.

Yeah.

And so, you know, that really had me second guessing, should I go down the route of academia? Because, you know, this is what it's going to be like, then I don't think I have the innate capability to beat that vicious cycle of you build something, it goes nowhere, but yet somehow you get back to working on something new. It's like insanity in my mind. And so I had become disillusioned and I was pretty much searching for like, okay, what's my purpose in life? It's like, do I sell my soul, go work at Facebook, optimize ads? A lot of Stanford CS kids do that, and I don't blame them. It's like, what are you going to do? But that's when my dad had the heart attack. My mom rushed into the ER. The physician in the ER ordered the cardiac CT and the cardiologist gave me a call from the other side of the world, like saying like, hey, your dad's calcium score came out to be 139. 100 is like high risk. And like, you know, I'm like, I like, it is such a coincidence that I speak the same language. I understand exactly what you're saying because I spent a year of my life building that algorithm, right? So that's when it all started. It's like, that's how I got into building Bunker Hill. It's like, hey, this is not the last time someone's going to have a hard time getting their innovations to the market. And so, yeah.

Yeah. Well, tell us about the name. How, what's behind the name?

Um, could we play a quick game maybe?

Of course. Of course. Okay.

So I'll give you 3 stories, very quick ones, and you tell me which one's right. So obviously there's the Battle of Bunker Hill. It's the little guys against the big guys. In this case, the little guys are the researchers who are building all these wonderful AI use cases. And the big guys are like the healthcare monolith that is like providing you friction and things like that. So that's the first story. The second story is if you drive up from Palo Alto to San Francisco, which is what we were doing, you would come across this road called Bunker Hill Road. And so we were like, hey, we need a name for the company. And so we named it that. The third story is around a TV show called Pure Genius. I have yet to meet anyone who has watched that TV show apart from me thus far. Maybe you will be the first or your viewers might be the first, but it was centered around a hospital named Bunker Hill where research that was done in the morning was used in clinical practice in the afternoon. And we thought that, that speed of clinical translation was critical. So which one, which story do you think is the correct one?

I'm going with the, the one you just told, the last one.

The last one? Well, you were right. You're right, actually.

Wow.

That's, that's one of the few times people have guessed that to be the right one. Most people go for the second one, uh, the one where you're just driving up the road and you see—

Yeah. Whether that's too random. The first one's very plausible because that makes total sense.

But then I'm not from America, so, you know.

Right. Yeah. You have to know your history. It's very different for you. But, um, no, I, I, I like that. That's the way it should be. Absolutely. Not this, uh, you know, back in the day it was 17 years bench to bedside, you know?

Yeah.

The thing was discovered before it got into practice. I think it's about half of that now, but it's like way too long, especially in this day and age technology.

Yeah. I, I don't know much about pharmaceuticals that takes that long, but you know, for software. Yeah, you've been a CIO, you know what I'm talking about. It's like, why are we treating software the same way drugs are being treated? Let's capitalize on the fact that software is software.

Yeah. Yeah, it's frustrating. And that's one of the reasons we started Digital Voices, our podcast, and some other things that we do. Yeah, tell us some key outcomes. So you've been in this now for a few years, 6 years, right?

Yes, we've been actually in this for quite some time. We had a pretty slow start. Our first, we spent the first couple of years trying to basically create this consortium of academic centers that would let, would allow researchers within that consortium to share data with each other, share algorithms with each other. So Cleveland Clinic's part of that consortium, Stanford, UCSF, and 25 other health systems are a part of that consortium. So it's been very cool to see someone who has an idea be able to get collaboration in place, data in place for that collaboration within like 2, 3 days of embarking on that as opposed to a year as it would previously take. What I'm most proud of is what happened after we actually built that consortium, which very fortunately coincided with LLMs becoming large language models becoming a thing. So since then, We've been able to get so many use cases out the door. We have health systems using us for over 15 different use cases across different departments, some for clinical use cases, some for administrative use cases. I'm obviously more proud of the clinical ones because they're cooler. So like with this calcium scoring algorithm, for example, the one that I had worked on, we actually brought it to life and we are deployed at several health systems now. At one of them, we Found this police officer who got flagged by our system. He got a snail mail. Um, he went back to the hospital, he saw a cardiologist for the first time, and turns out his younger brother recently had a heart attack.

Hmm.

He has a long family history of heart disease, and guess what? Like, he himself was having chest pain, but he had dismissed it as heartburn. So we found this patient such at a point in time, you know, he didn't get through the stress tests and the diagnostic CAT successfully. So then he ended up getting a triple vessel bypass surgery and the physicians on the other side, they were pretty confident that we likely prevented a heart attack here that Had it not been for this kind of solution, then he likely would've had a heart attack in the next couple of months. And who knows if it would've been fatal or not, but it's stories like that. You know, stats say one thing, like numbers say one thing, but the stories are what powers us so much, you know?

Nish, you know, a lot of the audience are hospital administrators. If you were CEO of Cleveland Clinic, what would you do to sort of speed up this whole, it's very time-consuming, slow, and all these great ideas and great technologies, AI, you know, and we could be saving more lives. What would you do differently?

I can only talk about the software aspect of things and AI as it pertains to software. On that front, I think I would try to come up with some form of governance structure inside the hospital. Where you can rapidly fail with your use cases, with your AI, quickly catch it and change it. Like if you make the cost of failure or cost of you try something and it doesn't work go as close to zero as possible, whether that's monetary cost, personnel cost, clinical cost, what have you. If you create a low-risk environment, you can iterate very quickly. Like, it is pointless to have a 5-year plan right now, or even a 1-year plan in AI. You just have to be like, how can we be the most nimble and most agile? And I know that's usually not within the DNA of a large organization, like a health system, but I think there are ways that you can do that for AI and software in a specific way, where how can you create the most low-risk environment where you can rapidly test something? And if it doesn't work, who who cares? Just move on to the next thing. Just, yeah, sandboxes, governance structures, platforms, partnerships, working very closely with vendors likely all contribute to it, but it does require like some kind of like shift in how you think about these things.

Yeah. You know, a lot of the individuals who sort of are responsible for those areas today come from a more traditional background and mindset. And I think that's one of the problems is that they will have this structure that's worked for them, you know, 20 years ago, and they're trying to impose that same structure on modern software, modern capabilities. And it, it, it's, you can't, it does, it doesn't work like that.

I'm not advocating for like the move fast, break things, you know, thing from Silicon Valley. I'm thinking, I'm talking much more of Let's say like, why do we need 25 meetings spaced out one month at a time to discuss something that like helps you automate registry submissions? You know, it's an administrative task. One could even argue that the standard of care is pretty, pretty low today. Not even going that deep into it, it's like, you know, just what's the harm? Like, why are we spending so many cycles trying to evaluate something? something like that. So it's things like that. It's like there are low-risk things that are occurring at a high enough frequency that you can quickly iterate and catch mistakes if there are mistakes. Again, I'm also not advocating for having lower tolerance, or I guess higher tolerance, sorry. It's just more like create an environment that can help you iterate very quickly.

No, I am actually for both the things that you talked about. I speak to boards quite a bit. I was just with several very high-profile academic medical centers on the East Coast. That's what I told the boards, you know, you gotta change, shift the mindset completely. You've gotta increase your risk capability. And there's safe ways of doing it, Nisha. You already mentioned what is sandbox, digital twins. There's all sorts of things. But if, you know, I make fun of HR structures as well, where we still operate modern companies, but modern companies like Silicon Valley companies don't operate this way, but hospitals still operate in the same way that The same structures that when they built the pyramids, right? You have one person, they've got a span of control of 5 to 10, and they've got 5 to 10, and these big monolithic— of course nothing can get done. You got to think more agile. And so until you do that, you can't really embrace modernity and some of these tools.

So I don't— it's like if you asked me what I would do if I was a CEO of a hospital, And I would imagine that the CEO of a hospital has bigger fish to fry in today's environment with all the budget cuts and things like that, that they can't be the one to spearhead a lot of the AI initiatives. So another aspect is I think if you find a trusted person to be your, quote unquote, your chief AI officer, your AI czar, whatever you want to call them, to be like the trusted person, like, hey, I will focus on putting out the fires. and making sure that we have a tomorrow while you go focus on the vision and like what it could be. I think that could potentially work, maybe. Again, I don't have— I've obviously never been a hospital CEO before, so I'm definitely speaking way out of turn, but I think—

No, no, no, no, no, no. It's good to have your insight. It's not out of turn at all. It's good to have your perspective and insight. That's why I asked the question. But yeah, but again, we have created these AI positions, but a lot of them, not all of them, there's a lot of great ones out there. But a lot of them, they were CIO and with old thinking, old structure, and now they added AI to the title. So it's the same, you get the same nonsense, you know, and it requires, and it's not that a person who's been doing a job for 20 years a certain way can't become that, but that's why you need to make sure whoever you hire, whether it's someone external, someone younger, older, whatever, they just have to have this mindset of, you know, being agile and being quick and And moving things along, but yeah, not tolerating sort of the lack of progress we've had in healthcare for so long. They need to be frustrated.

Yeah, I don't think that this technology's forgiving enough where one could just say, hey, we'll just wait for the dust to settle and we'll figure out what to do next because it might be too late. And especially in competitive places where There are many health systems competing for market share. I would imagine that it is very possible for one health system to lap over the others if one embraces it faster.

You've given 3 or 4 great ideas. And before we leave Bunker Hill, I want to ask you a little bit about leadership. Before we leave Bunker Hill, what's next? Like, how do people find you and what's next for Bunker Hill?

So we have a website called bunkerhillhealth.com. You can also reach out to me directly. It's my first name, nish@bunkerhillhealth.com. I'm always interested in talking to health system partners who are excited about advancing this field forward. We are a very collaborative company. A place like Cleveland Clinic not only is a customer of ours, but is also, for example, contributing their own algorithms to our platform. They're helping work with other institutions through our consortium. Like it's really a wonderful collaborative where it's not a vendor-customer partnership, but a lot more of, hey, we are a true partner. And so I'm very excited about extending those kinds of relationships and growing those to other health systems as well.

So what are one or two key things you've learned along your journey in terms of leadership?

Difficult conversations don't become easier with time. You know, I'm surprised as to how much we as humans Lean on trading. Like we'll be fine with a, like a lot of chronic pain, but not be okay with a higher order of acute pain. We'll avoid that. And this is like, I've learned this at work, but I've then realized that it applies to every facet of your life, whether that's a difficult conversation with your partner, with your parents, with your siblings or what have you. And it's just like difficult conversations don't get easier over time. And so better just get it out of the way. And so I've found doing that typically leads to faster resolutions and fewer things becoming problems later down the road. So that's one thing that I've tried to implement very significantly and I've learned that throughout my leadership career.

Yeah, no, those are good things. What about, what do you do when you feel like your creativity stifled a little bit? You're thinking about what to do, but everyone runs into that brick wall. What do you like to do to just get out and clear your mind?

I go on runs and during those runs I like to daydream a lot. It's like, what could you do? You know, it's like, for example, a quick tangent, but we have tried to impose how we should bring a drug to market onto AI. But if AI delivers its problem in drug discovery, what if we imposed how we bring software to market onto drugs? Again, like, you know, we could have like a 5-hour discussion around this, but it becomes very cool when you just start like daydreaming about if technology works and you could get the society to behave herd optimally, what could be different? And it's very fun, you know, then it's like, oh, The opportunities are boundless, like where the world could be is very, it's still very far away. So even if we had a setback recently, it's like, there's just so much to be excited about still. And then, so I typically like to do that. I found it that I need to be motivated in order to do creative work. I can't just force myself into that.

Yeah, I'm gonna carve out, you know, 5 to 6 AM to be creative. Yeah. Hey, Nish, this has been a fascinating discussion. We talked everything about like where you grew up and how you grew up, which is pretty cool. And then some of the songs on your playlist, like Fun. I haven't heard that song in a long time. Now it's actually going through my mind right now, especially the beginning part with the acoustic or what's it called? No music, just the vocals.

The vocals, yeah.

Yeah. And yeah, it's a great start to a song. And then we talked about Bunker Hill and a lot of the different things that you're doing. We talked About leadership, gave a lot of great ideas, like what you would do if you were a hospital CEO. What did we miss, or is there anything you want to double down on? I'll give you the last word.

I think one thing that we didn't touch on is that today it seems like companies like ours, or not like ours, but companies in general seem to want to pick a lane. Like we are either a clinical AI company or an administrative or an operational AI company. And I find that to be kind of weird in today's day and age. Like, wouldn't you love if like you could take care of a patient's journey throughout, like start to finish? Like find a patient who is currently undiagnosed, get them triaged on a referral queue. When they do get to see a physician, make sure that the physician is equipped with all the information they need very, very quickly. And then when the physician, for example, prescribes a medication, how do you make sure that the prior authorization is done in an agile way? And once that is done, how do you ensure that the patient is properly put or not put as a part of a registry to monitor downstream outcomes? It's like, I don't see the reason to believe in this false dichotomy. I think we could potentially Have companies doing both. So that's one thing. Yeah.

Oh, that's awesome. Nish, we'll check back with you, see how you're doing. But yeah, we'll put all your information that you talked about on how to reach you and Bunker Hill in the show notes. Thank you so much for being our guest.

Thank you for having me. This was really fun. Thank you for listening to Digital Voices. We hope today's conversation sparked ideas, reflection, and connection. Subscribe on YouTube, Apple, and Spotify Podcasts so you don't miss an episode.