¶ Welcome & Guest Background
Welcome to the ThinkData podcast brought to you in partnership with DataWorks. If you want to stay up to date with the latest breakthroughs and trends in the world of data and artificial intelligence, And if you're curious about some of the strategies that companies and founders use to launch data and AI products, then you're in the right place. Our aim is to bring together a diverse lineup of fantastic guests.
from the founders through to accomplished leaders and product owners at some of the most fascinating data and AI companies worldwide. They will each offer you their own unique insight into what it takes to launch and scale a great data business. Thanks for tuning in and I hope you enjoy the episode.
Welcome to the ThinkData podcast in partnership with DataWorks. And today I'm joined by Matt Vale. He is the Chief Technology Officer at DataVant. They are a Series B healthcare technology company that operates the largest health data exchange. in the United States.
It's a really interesting product, and I'm looking forward to kind of digging into this. But for the listeners here, their platform helps healthcare organizations securely share data, which in turn kind of improves care, advances research, and ultimately... drives better patient outcomes. Matt, it's really good to have you on.
kind of had a quick conf lab beforehand but do you know you've got a really interesting background obviously you're an entrepreneur yourself exited that business a number of years back and you joined DataVant just over five years ago but I'd love for you to kind of
bring us up to speed with kind of who Matt is and kind of what ultimately brought you to DataVance. Yeah, thanks. Thanks for having me, Alex. Really excited to be here. And, you know, my background, I've always been really passionate about. personal health and early on about helping people live healthier lives and improving public health. And so my training originally is in biomedical engineering.
And for a long time, I thought that I wanted to practice medicine and was really fortunate to have the first few years of my career to spend them writing software in clinical research. So I was doing signal processing type work for stroke research, epileptic research in St. Louis, and continue to be really passionate about the life sciences and the impact on. healthcare, but found that what I really loved doing day to day was building products.
And so I took that skill and I went into industry, into the life sciences and started building software products for life sciences companies. Did that for several years before founding my own tech company. down in Los Angeles, where we were focused on agricultural life sciences and food tech and spent about three and a half years doing that. Exited that business and moved up to the Bay Area, joined that event, as you said, about five and a half years ago.
When I joined the team was around 25 people. I was a seventh or eighth person on the technical team when we all reported to the founder and we're all sort of wearing all the hats and doing everything. engineering, technology, product development. And it's been an amazing journey over the last five and a half years. We've grown the business quite a lot and the team as well. And I've now been serving as our...
CTO for almost two years now, leading the technology team here at DataVamp. Interesting journey. Yeah, so obviously, you're techie over to your heart. You know, I won't go into this too much, but it was interesting looking into your background, that kind of vertical farm and the kind of, you know, the agritech startup to then move into what you originally said, kind of a passion of yours. How did you kind of manage that transition from...
¶ Transitioning from Founder to CTO
founder, kind of master your own destiny to, in a way, effectively working for someone? How did you manage that transition? Yeah, so I think I am most motivated by... having a big impact wherever I'm at. And so as a founder, it's fun because there is no one in finance, accounting, HR, there's no one in product management or, you know, any of the functional areas.
wear all the hats. And by doing that, you have a big impact, right? You're sort of building the business from scratch. But at Datavant, I feel I was very fortunate when I joined, we were still small enough that there were... a hundred different roles to play and gaps to fill. And our founder, Anish, who was my boss for several years is really...
a fantastic mentor. And I really appreciate the way that he gave all of us autonomy. And so it was honestly a relatively easy transition. I came to DataVance and I... just started finding the ways that I could have an impact. And on a lot of days that meant I was writing software myself directly. Other times it meant I was mentoring some of the earlier career engineers on the team. A lot of days I spent.
doing essentially product management when we didn't really have that function or we had one person playing that role. And so while it was definitely a shift, I think I was fortunate that The culture at DataVan really embraces autonomy and big scopes of ownership. And so the shift was really more in the exact work that I was doing and less so in the...
the ethos or the sort of motivation for me again is really around impact and also around learning and growth. So I think I was fortunate that Datavan had that kind of a culture early on. It's interesting, isn't it? As you said, a founder yourself, do you think that kind of...
founding mindset, the kind of hustle mentality, the find out for yourself, just kind of, it did help you when you made that transition into a, albeit a very early stage business, but did you, that just helped you get ahead quickly? Yeah, absolutely. And I think it still does today. I think the biggest thing about the founder mentality, in my opinion, is there's really this bias for action that you sort of appreciate.
Literally no one knows what they're doing and there's no right answer. And the worst thing you can do is fail to decide and take action. And so I think. moving forward, even in the face of this uncertainty, that to me is a huge part of what it means to be a founder. And I think that served me well early on here. But also now, as we're a bigger organization, I think we...
most companies, as they get larger, they have this risk of paralysis through analysis. And I think that that's one of my strengths is I'm comfortable with the idea that we're... literally never going to know all the data. And even if we did, there is no objectively right answer. So better to move forward in some way today and learn from the actions and then... iterate and take the feedback and improve that to wait for more information.
¶ Understanding Datavant's Health Data Exchange
Yeah, I love that. And I'm keen to touch on databank because I think that's obviously the whole premise of this discussion. You know, as I mentioned at the top, running the largest data exchange, obviously, healthcare data exchange in the US. For people that, you know... in the US, obviously they'll understand what that means in practice, but obviously with international listeners, what is DataVan and what problem is it ultimately solving for that healthcare space? Yeah.
Great question. So the way I would describe Datavant is as the health data platform for all healthcare. And so we're managing the secure and compliant exchange of health data. across the US and then in some areas in Europe, UK, actually at scale. And so we're solving this problem of data fragmentation across healthcare where...
Different countries obviously have different levels of this challenge, but I can speak to the U.S. It's easy for most folks to understand here. Every health system operates their own electronic health record. And those EHRs are all independent. And so you can think of it as a dozen databases being managed by each of these health systems. But of course, as people, we go to many health systems.
And I've moved with my family six times in the last as many years. We've gone to probably two or three times that many health systems. And without Datavant, those health systems are typically exchanging data. via paper, via fax, literally sometimes with a CD or a phone call. And with that event, what we allow them to do is structure the data and then exchange it. in a usable, actionable format so that when I go to a new hospital, my provider there has my records from... Ideally, the vision is...
all of my previous encounters at every health system I've ever been to. In somewhere like the UK with the NHS, it's a little bit simpler, but still. There are different implementations of the data model. There's unstructured data that isn't really actionable in its current format. And so there's still a lot of value to be had from connecting and structuring those different.
disparate data sets across those different hospitals. Yeah, it's really interesting. I'll start talking from the UK, but obviously the NHS is obviously one big entity, but obviously for the listeners here in the UK. The U.S. is the opposite. You know, it's a huge trillion dollar business and everyone operates as their own private businesses. And then obviously trying to, which doesn't help you as the patient, if you're ultimately moving from, say, Arizona to California, you think, well.
I need to make sure my GP and my physician has the right notes. And I'm interested to touch on, you know, we're talking about the data platform, data management, but within the business.
¶ AI and Data Quality in Healthcare
From an advanced analytics and data and AI standpoint, you know, as the chief technology officer, how focused are you on artificial intelligence to kind of do a lot of that heavy lifting for you? Yeah, great, great question. And definitely AI is very top of mind for us as it's, of course, in vogue everywhere in the industry.
The way we think about AI is a little bit different from a lot of the other companies that are hoping to do AI in healthcare today. Our approach to artificial intelligence really starts with the view that... healthcare has this data fragmentation problem and the data remains disparate and unstructured. A lot of these legacy paper-driven workflows, they also...
perpetuate this data siloing. And as a result, it makes it really, really difficult to develop a comprehensive and current view of patient health. And all of that is essential. to AI because at the end of the day, the models that we can train are only as good as the input data that we have for them. And I think while there's been some development in
novel algorithms. A lot of that happened a decade or more ago. And the major advances recently have really been about increased volume of data, increased quality of data. and increased training capacity or GPU resource usage. And so we're really approaching it as in order to build effective, secure, safe, compliant AI.
in healthcare, you have to go upstream. You have to centralize the data. You have to structure and standardize the data model. You have to take the unstructured text that is over 80% of the clinical data in healthcare today, and you have to make it usable by... by AI models and obviously with large language models, that's that's easier than ever. And that's really exciting to us. But if we don't take the steps upstream to clean the data and make it, as I said, ready for the AI, then.
you're just going to have garbage in and garbage out. So we spend the majority of our time thinking about bringing the data together and getting it into a format that is ready, ideal for the models. And then we do apply AI downstream to the way we describe it is to make the data actionable for our customers. And a lot of times what that means is.
abstracting particular structured fields out of unstructured clinical text. That's sort of the main technical workflow that we're doing with many use cases.
across life sciences partners, healthcare providers, and payers. Interesting. Yeah, you can be right about that kind of... kind of everyone thinks ai is the kind of silver bullet that's going to solve the problem but what they forget is you know if your organizational data structure and data warehouse or whatever however you're running your data environment you know if that data ultimately is is not accurate then no amount of you know uh
models is going to basically give you the insights that you need anywhere, at least accurate insights. It's going to give you something, but it's ultimately going to throw up some...
¶ Handling Identified and De-identified Data
curveballs. And you touched on structured and unstructured data, which is an interesting point because you also handle identified and de-identified data. And for people listening here, there's obviously... You've got the privacy side of things, so patient privacy, where it's like research, analytics, and then you've also got the identified data side of things. It's obviously really important for treatment.
How does DataVant, from a technology standpoint, handle that sheer volume but equally know which is which and make sure both are secure? Yeah. So the... On the scale side of things, we really leverage partners like AWS, especially Databricks. In 2025, those cloud computing partners are really at a scale where we leverage them to help us handle that data at scale. And we're talking about...
hundreds of terabytes and tens of millions of records every year, millions and millions of pages of clinical documentation. But with the scale of AWS today, The actual infrastructure is available for us, which is amazing. And I think it's one of the reasons that this is the right timing. Certainly 20 years ago, trying to handle this skill of data. distributed with disaster recovery and security and compliance the way that we
we hold to really the highest standards in the industry would have been extremely difficult, extremely expensive. And so I do think that is part of why now is absolutely the right time. A lot of those lower level problems have been solved. As far as the de-identified and identified data, we are extremely thoughtful and strict about segregating those two, and they live in different environments.
and are under quite different role-based access controls and security compliance requirements where the de-identified data. It is still restricted access, so on an as-needed basis. However, that as-needed basis is a larger group of folks who can access it within the Datavan engineering team product team.
On the identified side, it's a relatively limited subset of folks who can access those records. And in most cases, people don't need to touch the identified data to do their day-to-day job. In some cases, we do apply. We call them obfuscation tools, but methods of de-identifying the data in our production environments. So that, for example, as an engineer, if I need to go in and work on a production incident, I can...
debug something, access some data, but do so in a way where I'm still not observing the identified data. All of those things help us to minimize access. and therefore reduce any risk of disclosing that identified data in a way that it shouldn't be. But it really starts all the way back. at the initial design of our products. Every engineer at Datavant goes through security compliance training from day one. It is part of every engineer's job, something we take extremely, extremely seriously.
amazing chief information security officer or CISO with a very strong security team that helps to educate and train not just our engineers, but everyone on the team so that we're really developing. security and privacy into our products from day one and really utilizing a defense in-depth approach. So multiple different methods of securing the data, protecting the data and the environment.
from all the way from the user experience down to simple things like encrypting the data at rest and in transit, which should be standard practice everywhere at this point. Yeah, it's really interesting. You touched on something earlier about when you're moving.
say, state or from east of the West Coast. And back in the day, you may have your healthcare fire on a CD or it's being faxed. What you basically do is far more secure than someone losing your health record on a CD, for example. So I think it's, you know.
There's always that reluctance and nervousness, isn't there, when people are, when it's your most personal of data being held, not held by a third party, but being managed and transitioned by a third party. So is a lot of your work with the kind of...
those third parties, obviously the health systems themselves around getting their own house in order first before you look to kind of move that data. So they need to have a very secure environment before you look to move things. Yeah. I mean, you're actually right. And we take it. super seriously, it is arguably the most sensitive data in the world. I also agree with you when I look at the volume of medical records that hospitals still move.
via paper, via fax. Yeah, it is boxes and boxes of it every year. And to me, it's terrifying. It is less secure, to your point. But, you know, today, most of the health systems that... work with, or I guess I should say, we are fortunate to work with a lot of the largest health systems. So I think something like 75 of the top 100 by patient volume. And those hospitals, they take their data security and privacy.
equally seriously. And so the IT teams, the engineering teams at those health systems are first class and really focused on. protecting their data. So we don't usually have the experience where we're setting that up. They've, for the most part, already implemented their own EHR and their own security protocols that are kind of best in practice. in class, recommended often by the HRs. Where we, I think, provide a lot of value on the security front is in the connection between the systems.
whether it's at individual health system or multiple. So many health systems have a dozen different EHRs, inclusive of... You might have a separate practice management system for patient scheduling. You might have a separate Billy or RevCycle management system coding platform.
You probably have your employees' productivity tools. So you could, at a health system, have a dozen different applications that all hold some form of protected health information or PHI. And one of the value adds... that that event brings in is we're, we're stitching together these systems for you in a way that is HIPAA compliant, SOCQ. Hytrust, FedRAMP, all of the strongest compliance regimes and doing that really with security practices in mind.
So that is one area where even within a single health system, we do bring expertise in consolidating their tools and their systems. And then, of course, as we exchange data between health systems, we're doing that in that extremely secure way that, yeah, to your point, is definitely better than just sending a piece of mail.
I would say that's sort of where the biggest security advantage comes in is in being these secure pipes between the house systems. And also, you shouldn't forget between the 300 plus. real world data sources. So those are the de-identified data sources that we're also creating an exchange for with our de-identification. and linkage software products. Same story there, we're providing secure
and compliant mechanism for de-identifying that data so that it's no longer protected health information. It cannot be tied back to any individual and also then transmitting and sharing it securely with the partners. should have access. Comprehensive. Yeah, I think for anyone listening here, I think that's probably as tight as it can be. You know, I think it's great for yourself and your health system to take that so seriously.
¶ Challenges Scaling a Healthtech Company
I'm interested to touch on your journey. And obviously, we've all talked about you kind of joining DataBand. But as you join 2025 Folk, you know, early stage, you know, founder led, can you kind of over the last five, five and a half years, can you kind of... When you look back at a moment in time, what do you think some of the biggest things you've had to overcome in that journey is you've built not only the business,
to a certain point, but also the infrastructure and the technology that ultimately supports everything you've just said in terms of security, the HIPAA, everything. So talk to me about that because I'm keen to touch that because everyone thinks...
you know is everyone looks you're kind of roasting to spectacles don't they at your level think oh this must be you know relatively straightforward but is there a moment in time which you can touch on to think oh that was pretty tricky and this is how we overcame it yeah it's uh
Interesting question. I think there have been a few transitions from five, five and a half years ago until now. And I think each of them is sort of... may be represented as the team grows and you add another layer of expertise and leadership where at the beginning, there were eight of us writing software. And sort of your impact was like, how quickly can I produce high quality code? How quickly can I coach and mentor and support other engineers to produce high quality code? And how can I?
um, get to the right answer for sort of what we should be building by spending time with customers, with partners, with experts in the industry, et cetera. But it's really about like, how can I accelerate the tasks? Right. And then as you as we grew the team and grew the business, it really becomes more about how can I assemble the best, the strongest team, the best talent?
get the right people on the bus. And so I think that was a transition point to go from sort of this flat organization. Everyone is just contributing to the product hands on. to, okay, now we have sort of leadership structure. We've sort of made that transition, you know, sort of two additional times over the last several years. And most recently, when the founder left Niche.
And I took on this role. We've built a whole new leadership team made up of, and I'm amazed at the caliber of folks that are on our team today. But, you know, we have a technology senior leadership team made up of about 10 people, over half of whom have previously been chief technology officers. in some cases, chief product officers of health care companies around the country, very successful businesses. And so I think like zooming out and changing.
Software engineers will appreciate this. The level of abstraction at which you work, you're no longer working at that sort of like execute the task level of abstraction. But instead, it's now thinking about how do you assemble? new teams, entire new organizations within technology to go tackle the biggest problems in healthcare. And so the challenges now are really a lot about navigating.
the industry? Who are the customers? What are their biggest pain points? And what are the use cases and opportunities where we can deliver more value? And then really hiring. the right people, the most exceptional people, and assembling them into new organizations within technology that can go tackle some of these problems. So really like talent acquisition.
and understanding the landscape and the industry. So the challenges have definitely shifted dramatically over that time. And to your original question, it's hard to pinpoint a single moment in time, but maybe half a dozen. as these layers of abstraction increase. Yeah, it's interesting. Obviously, you touched on talent a few things there, a few times there in terms of your current setup, those kind of 10 folk who are very seasoned, very experienced.
¶ Hiring and Retaining Talent
You know, I see talent, both the hiring and retaining is kind of the lifeblood of any growing business. But when you look back to what you were hiring for... uh early on to versus to maybe what's changed more recently talk to me how they kind of approach that now because as you say people are still joining a journey but maybe it's a different type of journey so tell me through that yeah i think early on
And I think this is true of most startup companies. I think you're looking almost entirely for what I would describe as a general purpose athlete. So someone can play all positions on the field or wear all the hats. Someone who's happy to drop everything. be a product designer in Figma for the day or go talk to customers for a week and be the product manager or come back and then develop some new data model and build an API around it.
Early on, you need those people who will play all those positions and do so happily in an excited way. I think a lot of that is driven by curiosity and a desire to grow. outwards. And over time, I think you shift towards requiring folks who are more specialist. And we're at the scale where I would say we're somewhere in between.
And to sort of paint the other end of the extreme, if you look at, say, someone like Google, you know, they're going to employ dozens of principal level engineers who do nothing but think about how to optimize the search algorithm. and really like very, very targeted, you know, PhD level, postdoc level experience in that one field. And that's not who we are today. We're somewhere in the middle, but we are at the stage where...
we no longer need to hire engineers who want to go spend a week being a product designer, right? And the trade-off there is we then get to hire... folks who have a deeper experience, deeper domain expertise. We do look for folks now who have deep experience in healthcare, especially in provider. SaaS businesses, because I think those are...
They're quite rare and hard to find folks who've got hands-on experience building really high-quality SaaS products for providers. So that's an area where we do focus our talent acquisition. But also, for example, Five years ago, we never would have hired somebody who was an exclusive front-end expert. And now we hire many dedicated, front-end focused React. professionals who have built their career in TypeScript, React applications. And now we need that level of...
of expertise and specificity. And so I think that's been the biggest transition is sort of from generalist to not even specialist, but I would say someone more specialized at waste time. Yeah, it's interesting to touch that. I know you've, you know, AI is obviously front and center of a lot of discussions. And a lot of discussions I have on this actually to leaders like yourself is...
¶ Exciting Technology Trends in Healthcare
How is AI going to shape how technology functions? are built and i think in i think we're a little bit too early to make that decision now i think that whilst we're after disruption it also creates opportunity but from a technology standpoint you know just ecosystem if you look at it generally
If you look at healthcare specifically, what are you most excited about with regards to, it might not necessarily be AI, but if you look technology-wise, what are you most excited about? Yeah, a lot of things. I'll start maybe, I'm not sure if this is where you're headed with the question, but I think it's obviously a big topic. Will AI replace software engineers? And I think...
Usually it's presented with this kind of doom and gloom lens, but I think the reality is far more optimistic. I think there are a million times more software products to be built. than we have the bandwidth to build right now. And with the help of new AI tools, I think what we're going to see is actually that we have more people who are able.
to pursue those ideas, develop MVPs, launch products and see what creates value for folks. I don't think that that's ever going to replace the need for a software engineer, but I do think it will change. the skill set that's required in much the same way that, you know, 60, 70 years ago, everyone wrote in machine code. And then that got abstracted away into an assembly language.
which kind of abstracts away to these object-oriented C, then C++, Java-type languages, most recently Python. And writing software in Python is quite a different exercise than writing something in machine code. And like most of us who have come up through software engineering over the last two decades, like we would fail miserably to pay enough attention to the minute details of exactly.
what needs to be punched on the card and the machine code. But on the flip side, we're much more focused on what is the user experience? What is the customer problem we're actually trying to solve? And how can I do that most effectively? And I think we're just going to see that.
AI supporting developers continues that shift, that trend, where engineers need to become more and more customer aware, business aware, commercially focused, and understanding what is the actual use case that I'm solving. So I'm super excited about that. I think it brings the engineer closer to the customer, which for me is very fun and I think is a wonderful thing. So that's maybe the...
on that first half. But as far as the sort of new developments for data in healthcare, there are two things I'd love to mention. First one is not AI related, but... What we're seeing is that there is an increase in what I would describe as cloud-first data discovery. So historically, if you worked at a life sciences company or a provider, a payer, you would have a vendor.
who pitches you on a new data source and you would sort of let them tell you how to use it. And you go through this long procurement process and kind of figure out like, is this something that's worth me spending sometimes many millions of dollars on this data set? And now with new products from places like AWS, like clean rooms and some of their data marketplaces, we're seeing is more and more.
Researchers especially are going straight to these AWS marketplaces and others to discover the data that they can use. to solve problems, answer research questions and augment their clinical trials, all sorts of things like that. And so I think that is a super exciting development.
For folks who are familiar with online data sources like Kaggle, you can think of it as an extremely mature, well-vetted... sort of standardized data model that you can trust for clinical research or epidemiology or other sort of work like that. I think more and more data sources that are not included in that type of a marketplace or a clean room type situation are not going to get discovered. And I think what's exciting is that it means more researchers using more data.
to power more healthcare decisions, which ultimately will lead to better choices, better research, better progress in healthcare. So I think that's a really exciting development that we're seeing and I'm looking forward to it. But both participating in it as part of that event, but also seeing where the broader industry goes on that front. And then the last thing I'll share is I'm super excited about.
how we can apply some of the new large language models and other language AI to unstructured clinical data. As I said before, over 80% of clinical data is unstructured. It's free text written by a physician in the clinical notes. And most of that data goes... basically unused today, except for billing purposes. And I think that two, three years from now, all that data is going to be getting structured.
getting abstracted in a way that makes it usable for all sorts of super exciting things. I think five years from now, we'll be in a world where clinicians have decision support. for all of their patients real time based on 100%, not just of that patient's historical. clinical notes, but actually using models that are trained on the aggregate of a huge chunk of healthcare data across the world. As a result, physicians are going to be empowered to make better decisions.
help their patients with better care. So I think that's probably the most exciting development in healthcare, in my opinion. Yeah, it's fascinating. And ultimately, it's just going to lead to better patient care, better outcomes. That's exactly right.
¶ Improving Patient Care and Conclusion
And that's what we're here in the healthcare space. My day job is recruiting into health tech companies. And when you've got such a... clear vision and the benefit, you know, frontline care from an AI standpoint, I think we're a little way off that, but in terms of what you're doing, the workflow automation, the, you know, the assisting decisions, cleaning of data and actually making more sense of all of that data that's there.
is ultimately going to feed back to the patient, which, yeah, that's fascinating. Look, Matt, I've really enjoyed it. I think what you guys are doing at DataVan is... You know, no surprise, you're the leader in this space. I think in terms of technology, what you're doing and how you're building the business. Am I right in thinking if people are listening to this, they can go to the website? Can they see some of the use cases? Can they start to see, understand a bit more about the...
product if they're curious? Yeah, absolutely. They should be able to go to datavant.com and kind of read about the products and solutions that we offer to life sciences customers, healthcare providers, and health insurance companies. And then also we're actively hiring. So folks can go to the careers page on datavant.com and they can find many roles, especially in engineering and product. I'm looking for folks.
all different types of backgrounds, especially people who have experience in healthcare, but also folks who don't and have worked with big data and other sort of SaaS products and other technologies that we're hiring a lot. right now. And that's something I'm super excited. Hopefully, folks will hear this and come apply a data event. Exciting. Thanks, man. I appreciate it. And have a great weekend when it lands. Appreciate it. Awesome. Thanks, Alex. Good chatting with you. You too. Thank you.
Cheers. Bye.