Co-founder and CEO of Century Health, the New York company mining real-world evidence out of the messiest data in medicine: the doctor's note.
Somewhere in a hospital tonight, a clinician is typing a sentence that could change how a drug gets developed. It will describe a tumor's margins, a patient's response, a symptom that appeared three weeks before anyone expected it. Then that sentence will do what almost all clinical sentences do: nothing. It will sit in a field no algorithm can read, in a system no researcher can reach, useful to exactly one appointment and then effectively gone.
Vish Srivastava built a company around the refusal to accept that ending. As co-founder and CEO of Century Health, he runs a team whose entire job is to walk into the most valuable, least usable data in the world and walk out with evidence. Not surveys. Not new trials. The notes that already exist, already paid for, already written down during ordinary care.
His framing is almost stubbornly plain. "We can create dramatically better outcomes for patients," he has said, "if we can just make the data that's already collected from their care accessible and actionable." There is no moonshot in that sentence. There is a logistics problem dressed as a moral one, and Srivastava treats it as both.
The product at the center of Century Health is called CHARM, an AI abstraction engine. Strip away the category language and here is the trick: it reads unstructured clinical documents - a neurologist's free-text note, a pathology report, the running narrative of a patient's care - and it pulls out the specific clinical variables a researcher needs, structured and ready to analyze.
Two details separate this from a demo. The first is accuracy: Century reports 97% validated accuracy on that extraction, a number that matters enormously when the output is supposed to inform drug development rather than a marketing dashboard. The second is traceability. Every variable CHARM produces keeps a link back to the exact source document it came from. The machine does not just answer; it shows its work. In a field where a wrong data point can quietly poison a study, the receipt is the product.
Tomorrow's cures are hiding in today's patient records.
Century Health does not try to boil the ocean. It concentrates on a handful of therapeutic areas - gastroenterology, neurology, and rheumatology among them - and on disease territories where longitudinal records carry real signal. That focus is a strategy. Real-world evidence is only as good as the depth of the data behind it, and depth comes from going narrow before going wide.
The economics behind the bet are brutal and well known. Bringing a single drug to market can run on the order of $2.6 billion across roughly twelve years. Srivastava's argument is that a meaningful slice of that time and cost is spent re-collecting information patients already provided, simply because the original record was unreadable at scale. Make the existing record legible to a machine, and you do not just save money. You compress the distance between a patient's chart and the next person's cure.
For a company that lives in language models and clinical informatics, Century Health describes itself in resolutely human terms. Its team page calls the group "a passionate team of multidisciplinary builders," insists on a culture of people "first," and includes one line that doubles as a hiring filter: they don't tolerate jerks. The values read like a working contract - relentlessly mission-driven, humans first, bias to action, win with customers. For a roughly sixteen-person operation, that is less a poster on the wall and more a description of who survives the interview.
Srivastava's co-founder is Sanjay Hariharan, Century Health's CTO. Together they have built the kind of company that sounds modest in a sentence and turns out to be enormous in implication: take the data medicine already has, and finally let it speak.
In May 2026, Century Health announced an oversubscribed $5 million seed round led by Origin Ventures, with participation from InnovateHealth Ventures, 25madison, Next Play Ventures, and continuing backers 2048 Ventures and Alumni Ventures. Oversubscribed is the operative word. It is the market's way of saying the pitch survived contact with people who scrutinize healthcare claims for a living - and that the 97% number held up under questions.
What Srivastava is selling investors is not novelty. Plenty of companies promise AI for healthcare. What is rarer is the discipline of traceability, the refusal to hand a researcher a black box, the insistence that an extracted fact must always point back to where it was born. In an industry that has learned to distrust confident software, building trust into the architecture is the differentiator.
Listen to Srivastava and a pattern emerges. He does not talk like someone trying to disrupt medicine. He talks like someone trying to finish a sentence medicine started and abandoned. The data exists. The patients already gave it. The only thing missing was a way to read it at scale, accurately, with a paper trail. That is the entire enterprise, and its smallness is precisely why it might work.
If Century Health succeeds, the win will not arrive as a headline-grabbing breakthrough. It will arrive quietly, as a clinical trial that recruited faster because a registry already existed, as a treatment validated against a population nobody had to re-enroll, as a researcher who got an answer in weeks instead of years. The strange specific that proves the point is the one Srivastava keeps returning to: the most important sentence in healthcare is the one already written, sitting in a note, waiting for someone to come read it.
Reading: CHARM turns free-text clinical notes into structured, research-ready variables, and keeps a link back to the original document for each one.