The Power of Network

So I was talking to somebody — a machine-learning person who'd also done an MD-PhD, working on machine-learning applications in healthcare. We were chatting about ML product and engineering jobs in health tech, and at the end of the conversation, almost as an offhand comment, they said: "By the way, venture capital — I don't know if you've ever considered it, but for someone with your skill set it's actually an interesting career path. And there's a new fund, Northpond Ventures, that just opened up in Bethesda, down the street from you."

I knew very little about venture. I say this jokingly, but it's actually kind of serious: the extent of what I knew about venture was what I saw on the show Silicon Valley. In hindsight it turns out that's pretty accurate — but that really was all I knew. I didn't have a business or finance background, or really any interest. But I said, "Interesting, I'll take a look."

I looked up Northpond, and I looked up one of our co-founding partners, Mike Rubin, who's an MD-PhD as well. This was the real pivotal moment. I sent a cold reach-out without too much thought — essentially, "I'd love to chat and hear more about your career path. By the way, here's my background, similar to yours, MD-PhD." The critical piece, it turned out, was mentioning that my research was machine-learning focused, because Northpond was building a technology practice and looking for someone with exactly my background. Two days later, Mike responded to a cold reach-out from a stranger. We had a conversation about his career path, and he sold me on venture capital — why it's a fantastic place for someone with a technical, scientific bent.

About a week after that, I started a fellowship during my last year of med school, when I had some extra time to spare. It immediately clicked. Two weeks later I figured it was a decent shot this was what I'd end up doing. A month in, they made a full-time offer, and it was an easy decision to make the jump. So I kept working at Northpond through my last year of medical school, then joined full-time after graduating, a little over a year ago.

To summarize the take-home message — and I try to communicate this to anyone who'll listen — it's critically important to have five-year and ten-year plans. Not because you'll stick to them; you probably won't. I see them as important for having a higher-level logic or philosophy for how you spend your time, so you're not just wandering around aimlessly. But the reality is that serendipity governs everything. I really do believe this: life is a series of seemingly insignificant events that have massive downstream, life-altering outcomes. In pre-med and medical school there's this very tried-and-true path — pre-med to medical school, medical school to residency, residency to private practice. It's very well-trodden, and for many people that's the right decision. But it's important not to be too formulaic, and to keep an open mind to points of serendipity.

I do, to some degree. When I put this together in the Twitter thread, my advice was: always ask for advice, not for a job. It's a play on "ask for advice, not for money" — I forget who originally said it. If you're fundraising, for a fund or a startup, you could reach out to respected, trusted people and ask for their advice, and obviously in the back of their minds they're thinking, "Maybe I'd invest," trying to get to know you outside the explicit fundraising conversation.

For me it's a couple of things. People are more inclined to respond when it's not an explicit ask, because they may or may not have a job, and the point isn't for you to get a job right now — the opportunity could come a year down the line. But more importantly, I had no idea whether I even wanted to do venture. I was reaching out because it seemed interesting, trying to explore all possible opportunities. If I'd gone in thinking, "Venture — I've seen Silicon Valley, it seems stupid and not for me," I just wouldn't have done it. So don't go in with preconceived notions of what you may or may not want. Go in with beginner's mind, an open posture, and learn more — and opportunities may arise at the back end that you didn't anticipate.

I do. I'm a computer nerd, so it all goes back to AI. This is huge in reinforcement learning — DeepMind and others pioneered it with AlphaGo. There's 100% this explore-exploit trade-off. And I don't want to overstate exploring; you really have to get into the mindset of exploiting once you find something.

One piece of advice I often give around whether to do an MD-PhD versus something else is that it maintains optionality — you want to make the decision that opens the most doors, or closes the fewest. For training in your 20s, that makes a lot of sense. But if you're constantly exploring and never exploiting, just taking the path of least resistance that keeps the most doors open, that's a path to nowhere — an infinite number of forks in the road. So you need both. I often see people doing one or the other. In my world, in medicine and science, it's very much exploit: people may despise clinical medicine but stay the course anyway, because they're over-optimized on exploit. But it's a super useful mental framework, and I use it myself.

Another great point. Ultimately advice is super contextual and personal, and there's a tendency to over-index on any one piece of it — one reaches out to lots of people because one's procrastinating, or indecisive, whatever. But the key thing is these are all just data points. Take one example I discuss a lot: whether to do a residency after medical school. No one can answer that for you. It's incredibly personal — it's like asking, "Should I marry this person?" when the path is difficult, time-consuming and emotionally draining, as medicine is. It's important to talk to people; all of these things are data points. But at some point you have to make the decision yourself — integrate all those data points into a coherent picture that makes sense to you. If you find yourself having endless conversations, taking on everyone's advice, and your decision is just indexed to whatever the last conversation was, then it's time to choose a path and dive in headfirst.

For sure. This was one of the most popular points in that thread, which I found interesting. I'll get into the granular, but first, even before the tactical stuff and the systems: when you're building a network and being out in the ecosystem, it's super important not to be transactional. Be genuine. Be genuinely interested in people before you try to be interesting yourself. This is probably ingrained in me from growing up in the Midwest, where people are just friendly by nature. Be a good human being; always be thinking about what you can do to help, in addition to what others might offer you. In a relationship-driven business like venture, this is critical. Everyone knows that who you know matters, but every day, two years into this transition, I'm endlessly impressed by how true it is. Your reputation and the way you carry yourself in the ecosystem are hugely important. So be a good person, be genuine, help where you can, and don't only reach out when you need something. That's point number one — it's obvious, but it's the most critical piece. Forget the CRM systems: this is the one thing you have to get right, or none of the other stuff matters.

Beyond that, the systematic stuff really helps someone like me, coming from the PhD world. At the risk of oversharing, I'm an introvert-extrovert hybrid who skews introvert, so these systems help me stay on top of something I don't naturally gravitate to. I enjoy the relationship side of venture — the endless conversations — but having systems in place helps.

The way I do it is simple. People ask what software I use, but it doesn't really matter — I use Notion, but it could be pen and paper, Excel, whatever. The main point is: any time you meet someone worth staying in touch with, write their name down. There are people you'll immediately click with and stay friends with for life — you don't need a system for them. It's the people you chat with a few times a year where the system helps. So I write their name down and tag them with their interests, expertise or position — physician, scientist, radiology, neuro, payer if they work for a health plan — something that helps me index what this person is about.

Location is also critical, and this is a key lesson in the age of Zoom. Remote work has been incredible, especially for me, building a life in the DC area, which isn't a typical health-tech or biotech hub — it helps scale interactions. But there's no substitute for in-person interaction. So I build relationships in person and then maintain them remotely, and tagging where people are located is key. If I make a trip to Boston or the Bay Area, I can immediately see who I know there, and who I haven't spoken to in a while, and reach out to grab a coffee or a beer.

The last piece I really like is a date field — the last time you interacted with someone. You can do this formulaically in Excel or Notion, and it'll automatically flag when it's been three months, six months, a year — whatever you set — to remind you to reach out. And going back to not being transactional: if I talk to someone I know is interested in, say, obesity research, and a week or two ago I saw a New England Journal of Medicine paper on a new obesity drug with really impressive results, I immediately think of them — I remember that's what their PhD was in. They've probably seen it, but I send it anyway, in case they haven't. The point is: don't only reach out when you need something, or when it's "time." Reach out when you see something relevant to that person — or maybe you're on the board of a company that's hiring for a role they'd be great for. It's always this give-and-take, which is really critical when building networks and relationships.

It's fair criticism, but I don't see it that way. The fact that humans have terrible memories isn't a function of being disingenuous or not genuinely interested. So I'm glad you brought it up, because note-taking is another key piece — not just for networks and relationships, but for so much of investing. It's probably the most critical thing, because I'm by no means a super-memorizer.

Again the software doesn't matter, but one thing I've loved over the last couple of years is Obsidian. Roam Research is another. They're these network-based note-taking systems with a very flat hierarchy — you're not writing a note and thinking about where to file it; you just start writing and connect it to other notes and topics. It's very similar to how human memory actually works, neuroscientifically, so it vibes with the way my brain works.

The point is: I take so many calls in venture — pitch calls, follow-ups, ecosystem calls — you're never going to remember all of it, no matter how genuinely interested you are. So keeping notes means I know what we discussed before. When you and I catch up six months from now, I can pull up the note from our previous conversation while taking notes on the current one, and remember that you mentioned something I wanted to follow up on. It's super seamless.

And not to get too nerdy, but as augmented reality gets adopted — Apple's working on it, Facebook's working on it — I think this is the next iteration of general computing. You can imagine a future where you're wearing glasses with an AR interface that automatically recognizes a face and pulls up your previous notes, so you immediately remember you discussed X, Y and Z. I can see the dystopian view — that this boils everything down to computer code and transactions — but I also see the optimistic view, where it just augments our existing relationships.

I haven't read the Gary Vee book, but it's an interesting idea. The way I immediately contextualize it: this whole give-and-take trade-off presupposes that we, as humans, understand what we want. I think we very rarely understand what we want in the moment — and even more rarely can we predict what will be relevant to us in the future. So if you have to be utilitarian about it, the right way to think about it is: I have a bunch to offer. Everyone has to answer this personally, but for me — this has been true over the last two months because of the Twitter thread — I have a unique experience, having come from medicine and made this transition, which is incredibly difficult because there are so few resources in academia and medicine for thinking about alternative career paths.

The funny thing is, I wrote that thread to decrease the number of one-on-one advice conversations I'd have — and it's completely backfired. I'm half-joking, but I'm happy to do it, because I wish more people in my position had been able to have these conversations with me when I was starting out. So think about what you have to offer, get out into the world, and distribute that skill or knowledge. At some point in the future — and if you need to be selfish about it — there'll be something you can offer that person, or they can offer you. You may not be able to predict what, because who knows what any of us will be doing in five years; it's probably not what we think. So the default posture is always: what can I offer? That's the best way to ultimately get what you want, and also the best way to bootstrap this whole ecosystem into a positive-sum world.

Domain expertise is key. Caveat: I'm obviously colored by my own experience — my reality is an n of one, so I can't over-generalize. I've had a number of conversations with people early in their PhD, trying to pick a lab, who say, "What's hot in venture right now? In the latest biotech wave, what's most interesting — is it CRISPR, RNA editing? That's what I want to do my PhD in, so that in five years, when I come out, I'll be most sought-after." My position is that's a slightly dangerous way to think about it. One, who knows what'll be hot in five years — it's hard to predict. But more importantly, rather than chasing specific skills you think will be applicable, what you actually want to do is cool stuff — research and projects that are interesting because they're interesting, and nothing more.

By the way, if you plan to go the academic route — build a lab, chase NIH grants — none of this applies. But say you want to go into biotech or venture: what you really want is to work on things you love. For me, even though my project had, from my perspective, zero real-world application — I was never going to turn my sensory-substitution device into a company — the point was, number one, it was cool. It's something I can talk about to people who know nothing about neuroscience and they still find it interesting; it gives me a memorable story. But most importantly, it was something I loved to wake up and think about every day, and I built a bunch of transferable skills — programming, machine learning — that turned out to be relevant to venture, where I invest in lots of machine-learning-enabled companies.

So when developing domain expertise, it's not even really about the specific domain. It's about developing a name for yourself in something, and that something will yield generalizable skills — critical thinking, a scientific mind — that are applicable no matter what.

The last thing, which has guided how I've crafted my own trajectory: look at the intersections of fields that don't, at face value, have anything to do with each other. Some of the most interesting domain expertise comes from creating a totally new domain by combining two areas. For me it was machine learning and medicine, which isn't super novel any more, but the principle holds. I was never going to be a 10x Google engineer, top-1% ML practitioner, but I could be above average at that, and combine it with an above-average skill set in clinical medicine — and suddenly you're doing something no one else can. So find two different skill sets and combine them into a new domain. That's how you set yourself apart.

Writing skills — and communication skills more broadly — are critical. A few things. One, write early. I constantly hear people say, "I don't really have anything to write about, so I'll wait until some epiphany strikes." That's the wrong way to think about it. As others have said — this isn't my idea — you don't write what you think; you write to discover what you think. The number of times I've sat down to write something and realized, in the process, that there are three errors in how I've been thinking about it...

And for any PhD students listening: there's a sense in science that you should complete the entire experiment, the entire project, and then write it up. In my opinion, that's backwards. You should be writing at the very beginning. I changed this for my PhD after my experience in undergrad: once you start writing up the manuscript, you realize there are errors in how you've been thinking, and it changes how you do the project. That's generalizable beyond science.

The second thing is the easiest, because it doesn't require a blog or a podcast: just spend time proofreading and editing the stuff you send every day — emails, texts, Slacks. You fire off a three-paragraph email you wrote in five minutes and never re-read. The number of times I've stopped, gone back, and cut the word count by at least half while making it clearer, with five extra minutes of work — number one, it saves time for the person reading it. But more importantly, that's exactly what you're trying to practice: communicating complex ideas, distilling them to their fundamentals, doing it in fewer words. You can practice that every single day in the emails you send colleagues and the texts you send friends and family. It's probably one of the highest-yield things you can do.

It's so good — I'm glad you reminded me of that one. That's exactly it, and it's why it's hard to write succinctly: it takes time. And you're not always going to have the time.

It's something I haven't been super intentional about until the last few months. I've been on Twitter a long time, but only really active recently, so I'm trying to be more thoughtful. Let me be specific about personal brand — or really, writing and communicating content online, which indirectly builds brand. And by the way, plenty of people cringe when they hear "personal brand," and I'm the same — but the reality is it's important, despite the negative connotations.

For me there are two motivations. One goes back to writing to discover what you think: so much of what I do in venture is looking at new industries, markets and technologies where I know very little, trying to come to a point of view about where the future of X is going. Writing publicly is great, because you're soliciting feedback from people who are smarter and know more than you, and you're indirectly promoting your brand — people know you're thinking about this stuff.

The second piece is specific to venture and early-stage investing, though it probably applies elsewhere. So much of what we do is enabled by occupying mind-share. The earliest-stage investing is very much about access — you know there's some capable, talented founder coming out of a top academic lab, everybody wants to invest in them, but they may not even know that person is spinning something up. Having the relationships means you know they're building the company, and it'll be funded before it's ever announced. One way to build those relationships is being out in the ecosystem, having one-on-one conversations — everyone should do that. But the most scalable way is to communicate your ideas, theses and value-add publicly. The fact that you write — whether a Substack post or a Twitter thread — helps founders find you, in addition to you finding them. I might have a thesis on the future of machine learning in healthcare, and there's some founder out there I've never met with a similar idea, looking for someone to invest, and they think, "I don't know this guy, but..."

You're right to point that out — it does contradict a lot of what I'm saying, and this is the tricky part. I'm by no means good at saying no and balancing my time; everyone struggles with this, and I 100% do too. Early in your career there's always this trade-off, where you want to — and in many cases should — say yes to a lot of things. I'm always skeptical of people who say, "Say no to everything." If I'd said no to everything from the start of my career, I'd be nowhere, and I think most people are the same. But it's a super tight balance.

The way I thought about it: in the first couple of years of your PhD, or the first year of medical school, there's a baseline of foundational knowledge you're acquiring. And this is one mistake I made — anything not related to my specific topic, machine learning in neuroscience applied to sensory processing, I was laser-focused away from. Back in medical school, embryology, neurodevelopment, all these other areas — I did the bare minimum to get by. In hindsight I regret that, because a lot of that knowledge is actually relevant to what I do in venture now. I have a PhD in neuroscience; I should know that stuff. But six years ago, during my PhD, I had no idea this was what I'd be doing — I barely knew what venture capital was. That's why I hesitate to over-generalize.

You're there to learn foundational knowledge, and first and foremost to figure out: do I want to do a residency? Do I want to do a postdoc? If you're too oversubscribed and distracted, you won't get the full immersion needed to answer that question, or all the skills you need to move beyond graduation. But it's tricky. VC is hard to get into, so these VC fellowships are fantastic opportunities and great ways to gain exposure and potentially find a full-time role. There are other things you can do early in medical school, and being on Twitter is pretty context-dependent. I just think it's important to be very thoughtful about how you spend your time in your PhD and MD, and to have a high bar for other activities you take on — not never do it, but if it takes ten hours a week, it may or may not be a waste of time. Try it, but be deliberate about it.

There are a bunch of ways I could answer that. This will probably be the less juicy take — I have plenty of interesting takes on the future of healthcare, most of which will probably be wrong, but are interesting nonetheless. But there's one in particular that's instructive, in the vein of everything else we've discussed. Let me talk about machine learning and healthcare, and how my investment thesis has changed as I've gone from a more technical background to a much more business- and finance-minded investor over the last two years.

Right before I moved to venture, I was doing radiology-AI research — developing novel convolutional neural networks to detect things on chest X-ray, and thinking through methodological choices to get the most generalizable model across sites, scanners and demographics. Important work for the technical feasibility of ML in radiology. Then I moved to venture and naively thought, "Investing is simple: I'll find the radiology-AI company with the best algorithm, the most differentiated, highest-quality training data, the highest AUC, sensitivity and specificity. I'll diligence the algorithm." And it's so much more than that.

I wouldn't call this contrarian, but there's still a sense in the healthcare-AI space that the reason we don't have widely deployed healthcare AI is technical. The reason that, if I fell and hit my head and went to the hospital, I wouldn't actually get an algorithm detecting some pathology on my scan — it's not a technical question. Not to understate the importance of pushing the boundaries on generalizability, but I'd argue we already have the technical capability to derive clinical utility from these algorithms, and we're still not doing it at scale. So why?

Two years ago I'd have said it was technical — "If we just eke out a few more percentage points of AUC, the era of ML in healthcare will be upon us." It's much more than that. Number one, there hasn't been a venture-scalable business model to support these startups. There have been success stories, and I'm still long-term bullish, but in the absence of reimbursement — there are only a few reimbursement codes for a few algorithms — it's not widely scalable, and that makes it hard to support the layering-on of capabilities, the product development, the whole business case that lets these companies scale and keep deploying.

The second piece is clinical adoption. If you've ever watched a radiologist practice, they operate at absolute peak efficiency, so even the tiniest bit of friction introduced into their workflow — via a clinical-decision-support algorithm telling them whether there's pneumonia on a chest X-ray — is going to doom it. Clinical adoption is finicky, and the sad reality of medicine is that better patient outcomes are sometimes an insufficient ROI to drive adoption.

Finally — and this is something I'm critically interested in — there's an under-researched area of interaction: algorithm plus physician is almost always going to be better than either in isolation. People ask, "Is the radiology algorithm superhuman?" But if you look at the papers, the comparison is often algorithm versus radiologist versus radiologist-plus-algorithm, and the combined ensemble usually outperforms either alone. There's a paper in npj Digital Medicine — last year, I believe — that looked at how you bias, or don't bias, a physician when you deliver an AI prediction for a chest X-ray. They delivered a positive or negative AI read to the physician — "this X-ray is normal" or "this has pneumonia" — and importantly, sometimes they gave the wrong read, to see whether they could bias the physician. With erroneous AI predictions, the physician's read became worse than a coin flip. So it's not as simple as delivering an AI read on top of the clinician's workflow and assuming they integrate it as one data point and reach their own decision. There are lots of cognitive biases — anyone who's read Danny Kahneman's work knows how hard this stuff is to overcome as a human decision-maker.

So it's a long tangent, but the main point is: I came into this thinking it was a technical challenge, and that as we built better AI we'd get more venture-scalable companies and more patients getting AI in their care. In reality, the technical challenges matter and we keep making progress, but it's so much more — the idiosyncrasies of human-computer interaction, the business model, the lack of reimbursement, the payer perspective, how this transitions from fee-for-service to value-based care. That's a lot of the transition I've made over the last two years: understanding that we invest in businesses, not technologies. The technology is necessary but not sufficient; ultimately it's the business case and the revenue that matter.

Every single day — it's incredible. And you're in the UK. The NHS — I don't have direct experience with it, so I'm speaking a bit out of my lane, but compared to the US, in terms of organizational and incentive structures, it's just radically simpler. I'm very active at Northpond — I make investments, but I also take board seats and work closely with companies on strategic decisions, product, business models. A couple of the companies I work with are value-based-care oriented, transitioning from fee-for-service — collecting revenue according to physician utilization — to value-based care, where a company gets an explicit amount of capital, uses it to take care of patients, and sometimes keeps the difference as revenue. So there's an incentive to take good care of patients and do it cost-effectively.

But as I get into the weeds — learning how payers think, how reimbursement works, how things go from fee-for-service to value-based care, and where physician incentives sit, often to deliver care and bill rather than to be cost-effective — it's incredibly complicated. Over the last couple of years it's been a really critical experiment. Everyone's aware of how much venture money has gone into digital health, but the reality is these companies, with super-talented management teams, aren't going in saying, "We're going to gut the American healthcare system and rebuild it from scratch." They're working within the existing ecosystem, and I think they'll succeed. If we were starting over, idealistically, we'd probably build it differently — and I don't claim to have the right answer. Everyone I've encountered has the best intentions; I don't think there are real bad actors. Healthcare is incredibly complicated. We made the best decisions we could over many decades, and now we have a system that's a little discombobulated. But there's a very real chance for innovation, and I'm long-term bullish. If we could do it again, we might do it differently, but there's still going to be huge opportunity for success and impact, despite the challenges we're all aware of.

A couple of things I try to channel that I think can really help people. One is that radical self-awareness is critical — being really honest with yourself about your skills and your weaknesses.

Everyone talks about "follow your passion" — figure out what you're passionate about and do that. That's important advice, and the ideal scenario is that what you're passionate about — what you could do for hours on end and not think of as work — is also the thing you're top-10%, top-1% at. For me, interestingly, that wasn't the case. The thing I was most interested in was coding. If I could pick the one thing I'd be best in the world at, it'd be software or machine-learning engineering. It's like snow skiing or water skiing for me: I could do it for twelve hours uninterrupted, in hyper-focus. But the radical transparency I had to accept was that I was never going to be a Yann LeCun or a Yoshua Bengio — I'm naming the godfathers of AI. I was nowhere near that level.

So I wanted to find the thing I was most uniquely suited to do that I also liked — and I don't think I was uniquely suited to sit behind a computer building machine-learning models, even though that would've been the most fun. I know it's a dirty truth, but that's the truth. Radical self-awareness matters, because there's often a gap between what people want to be and what they actually are — you'll think of yourself as one thing because you want to, and it isn't true. I think it was Richard Feynman: the lies we tell others pale in comparison to the lies we tell ourselves. Be honest with yourself.

The second piece is leading with action. This has got me in trouble at times, but asking for forgiveness rather than permission — within reason — is the way to go. At the end of the day, no one's going to look out for you except you, and nobody's going to tell you what to do most of the time. You have to figure out what makes sense to you, even if you're not — and often you aren't — an expert in that space. Don't overstep, don't do things totally out of your lane; there are points where you should ask permission. But the people who are comfortable with uncertainty lead with action and think in solutions, not problems. The solutions aren't always right — often they aren't — but you're always thinking with that flip-forward attitude. In many areas of life, and venture especially, that's absolutely critical. So get out of your comfort zone, lead with action, think in solutions, be thoughtful about it — that's one of the most critical meta, philosophical skills in this space.

I thought you might ask this — I've got my Kindle app open. A lot depends on what you ultimately do. For venture — and these may be too specific for this podcast, but regardless — The Power Law just came out, a great historical perspective on the space: where we've come from, where we're going. There's also The Business of Venture Capital and Venture Deals, two standard recommended readings for anyone interested in venture.

Other books are less immediately actionable but have really shaped the way my mind works. One is The Structure of Scientific Revolutions by Thomas Kuhn. There's a sense that science is a very objective discipline — which is true to a degree — but the reality is that science is practiced by scientists, who are humans subject to biases and herd psychology. Kuhn walks through the history of science and shows how this works; it's actually where the term "paradigm shift" originates. It has a specific meaning: there's some dogma accepted by all of science, then findings that don't fit, which often get suppressed or dismissed by the big names in the field. Over ten years you get more and more inconsistent findings, which start to overthrow the existing dogma, and then you have a paradigm shift to the next one, and it iterates. It's a book most scientists should read — it formulated a lot of how I think about scientific shifts and the history of science, and it even informed some of my investing. It's probably under-read in the space.