Why I Built Vitality AI Health

It started with seven browser tabs.

One night I sat down at my desk and counted them: Oura, Whoop, Apple Health, Cronometer, my lab portal, a supplement tracker, and a glucose log I was maintaining in a spreadsheet because nothing else captured what I needed. All open at once. All telling me something different. None of them aware the others existed.

I had good data. More data, probably, than most people have ever had about their own bodies. And I could not answer a simple question: am I actually getting healthier?

That particular night I had eaten a meal around 8 p.m. and checked my glucose about ninety minutes later. The number was higher than I expected for what I thought was a reasonable dinner. Fine. Data point noted. But then I woke up the next morning and looked at my Oura sleep score, which was lower than usual. I had no way, in any of those seven tabs, to draw a line between the two. Did the glucose spike disrupt my sleep? Did the sleep disruption come first and influence my glucose response? Was there a pattern, or just noise?

That was the moment I knew I had to build something. Not another tracker. A layer that thinks about the connections.

I had been tracking my health seriously for about three years before that moment. I knew the individual tools were good. Oura's sleep staging is genuinely excellent. Cronometer is the most accurate food logging platform I have found. The glucose monitor I was using gave me information my doctor had never seen in real time. But good data in separate silos is not the same as understanding. It is closer to having the right ingredients and no recipe.

Think of it this way: your body has a dashboard, the same way a car does. The problem is that right now, every gauge is on a different screen, maintained by a different company, and none of them talk to each other. You can look at your oil pressure. You can look at your engine temperature. But you cannot see them together, and you cannot ask the car what they mean in combination.

A 2025 survey conducted by Talker Research for MDLive found that the average American now uses six different health apps regularly, and one in five uses ten or more. Fifty-three percent of respondents said there are too many health apps to keep track of, and forty percent admitted they do not know how to use the apps they have to their best advantage. I was in that forty percent, and I was building a spreadsheet as a workaround.

The data fragmentation problem is not a personal failure. It is structural. The Office of the National Coordinator for Health Information Technology tracks interoperability across the healthcare system and the picture at the consumer level is far worse than at the hospital level: clinical records have been improving their ability to share data between institutions for years, but personal health data from wearables, nutrition apps, and continuous monitors exists almost entirely outside that system, with no standards for exchange and no infrastructure designed to connect it.

That gap is where Vitality AI Health lives.

I did not build this alone. My cofounder Venkata Ramana brought the engineering rigor that turned the idea from a whiteboard concept into a product. We spent a long time together, longer than felt comfortable at the time, asking what it would actually mean to synthesize health data rather than just aggregate it. Aggregation is easy. A well-designed dashboard that pulls your Oura data and your Cronometer data onto the same screen is not that hard to build. But putting seven tabs on one screen does not solve the problem. The question is whether the platform can reason about the relationships between the signals, surface what is actually relevant, and change its interpretation as your data changes. That requires a different kind of architecture, and it took real time to get it right.

What we built is a score, the VAI score, that reflects your overall health trajectory at a given moment. It pulls from your sleep, your activity, your nutrition, your biomarkers, and increasingly your lab data, and it surfaces what matters most for you to pay attention to right now. If you want to understand how that score is actually calculated, I wrote a separate piece that goes into the methodology: what your VAI score actually means. The short version is that it is not a fitness score. It is not a wellness score. It is closer to a question: based on everything you are measuring, what direction are you moving?

Longevity is the frame I keep coming back to. Are you aging faster than your chronological age suggests, or slower? Most people never get to ask that question with real data behind it. That is the problem I wanted to fix.

When you spend this much time thinking about health, you start to notice something uncomfortable. The people who most need good health information often have the least access to it. Not because the apps do not exist, but because the friction of managing six separate platforms, understanding conflicting signals, and translating data into decisions is a part-time job. That reality shaped some of our choices as a company. We fully funded a school through Pencils of Promise because we believe the same principle applies in education: access to the right information at the right time changes trajectories.

It felt like the right way to put stakes in the ground about what we are actually building toward.

We have been shipping fast. If you want to see where the product stands right now, the April 2026 update covers everything we released in the last quarter, including the Vaya Chat improvements that have gotten the most feedback from early users. Vaya Chat is the conversational layer inside the platform where you can ask questions about your own data, not generic health questions but specific ones tied to your patterns. It is the part of the product I use the most myself, and watching it get better has been one of the more satisfying parts of this year.

I get asked sometimes why I did not just build a better individual tracker. Deeper sleep analysis, or a more accurate food log, or a smarter glucose correlation tool. The answer is that the tracker problem is largely solved. There are excellent tools for each of those things. What is not solved is the layer above: the synthesis, the pattern recognition across modalities, the ability to look at your glucose and your sleep and your HRV together and say something coherent about what they mean.

That is the product I needed and could not find. So we built it.

If you are someone who already tracks your health seriously and feels like you are swimming in data without clear direction, I think you will recognize the problem I described at the beginning. The seven tabs are familiar. The question of what to do with all of it is exactly what we are working to answer. That is the version of the problem worth solving, and it is the one that gets me out of bed in the morning.

Kevin Amrelle is the founder of Vitality AI Health.