Your annual checkup lasts maybe fifteen minutes. The remaining 8,760 hours of the year? That's when things tend to go sideways.
Prana, a four-person team fresh out of Y Combinator's latest batch, is betting that gap—the vast space between doctor visits—is where most preventable health crises take root. The San Francisco startup launched in mid-February with what amounts to an always-on medical surveillance system: artificial intelligence that watches streams of data from wearables and medical records, looking for trouble. When it finds something, a human doctor steps in. Usually within an hour.
The pitch is straightforward, if ambitious. Connect your Fitbit, your medical history, your glucose monitor. Let the AI watch for what the founders call "clinical drift"—the slow-motion slide from healthy to not. Pay nothing unless you actually need to talk to a physician, which starts at $39 per visit. No subscription. No insurance required.
Whether this model works—clinically, economically, or both—is the kind of question that has vexed digital health companies for a decade now. Forward Health thought it had figured out automated primary care with its futuristic CarePods. It shut down last November. K Health tried the AI-plus-doctor formula but ultimately pivoted toward partnerships with traditional health systems rather than going it alone with consumers.
Prana is trying anyway.
An Architecture Built on Other People's Infrastructure
The technical setup is less a breakthrough than a clever stitching together of existing tools. Prana uses Fasten Connect to pull in electronic health records from more than 50,000 healthcare organizations—a sprawling network that, in theory, means your lab results from one hospital can talk to your prescription history from another. Junction's API handles the wearable side, piping in data from over 300 devices. Garmin, Oura Ring, Dexcom continuous glucose monitors. The works.
The AI layer runs on models from OpenAI, Google Cloud, and Anthropic. It's watching for patterns: elevated resting heart rate that won't quit, blood glucose trending upward, sleep data gone haywire. "We watch your vitals, so you don't have to," the company says on its site—a promise that sounds reassuring until you remember how many false alarms most wearables generate.
When something looks off, the system routes you to a licensed physician. That's the human safety net, and also the revenue model. The AI triage is free. You pay only if you escalate to an actual doctor visit.
It's a deliberate departure from competitors like K Health, which charges $29 to $99 monthly whether you use the service or not. Whether pay-per-visit economics actually pencil out better than subscriptions is, perhaps, the central question here. If patients rarely need doctor interventions, Prana makes little money. If they need them constantly, the model looks appealing—but might suggest the AI isn't filtering effectively.
The Team: One Doctor, Two Engineers, and a Deferred Med School Admission
CEO Meer Patel studied biomedical engineering at Johns Hopkins and was accepted to Brown Medical School. He deferred to build AI health tools instead—a choice that speaks to either remarkable confidence or a certain recklessness that often characterizes founders, depending on how this turns out. His previous projects include ICPredict, an AI clinical deterioration predictor, and work on saliva diagnostics.
CTO Vishvam Rawal comes from the quantitative finance world—Weiss, Barclays—where reliability matters and mistakes cost money fast. His background is applied mathematics from UC Berkeley. The third co-founder, Sanjit Menon, is finishing medical school and previously built an AI medical education startup that reached 400 paying customers before this venture.
Four people total. They're operating under a management services organization model, which means Prana builds the software while affiliated medical groups handle the actual diagnosis, prescriptions, and treatment. The legal structure is designed to keep the company on the right side of regulations around who can practice medicine.
The terms of service are blunt about this division: "PRANA AI IS IN NO WAY ENGAGED IN THE PRACTICE OF MEDICINE." The AI's output "may be wrong," it cautions, and should be verified by a licensed professional. Which is to say: the technology is a triage tool, not a replacement for clinical judgment.
Fair enough. But it raises the question of where, exactly, the value proposition lives. Is it convenience? Cost? The promise that someone—something—is watching even when you're not thinking about your health?
A Market That Has Seen This Before
Digital primary care is littered with cautionary tales. Forward Health's collapse is just the most recent. The company raised over $100 million to build sleek, AI-powered "CarePods" where patients could get scans and tests without a human doctor present. The model didn't scale. By November 2024, Forward was done.
K Health is still standing but has notably shifted strategy. Rather than chasing individual consumers with a pure direct-to-consumer play, it's embedding its AI tools within existing health systems—a tacit acknowledgment that going head-to-head with traditional care is harder than the pitch decks suggested.
There's a broader trend here, though, that works in Prana's favor. Continuous monitoring is no longer fringe. A recent review in the biosensors literature noted how wearable bioelectronics, when integrated with electronic health records, can catch deterioration patterns that episodic clinic visits miss entirely. Remote patient monitoring vendors have reframed their pitch around "between-visit risk"—the idea that most dangerous changes happen at home, not in the doctor's office.
Academic support doesn't always translate to commercial viability. But it helps.
Available Now, With Asterisks
Prana says it offers nationwide access to board-certified physicians, though the fine print notes that availability varies by jurisdiction. Illinois and Nevada have specific restrictions around mental and behavioral health queries. The platform is for users 18 and up. Emergency cases are explicitly out of scope—call 911, the disclaimers say, not us.
The infrastructure runs on AWS with TLS 1.2+ encryption for data in transit and AES-256 for data at rest. Business associate agreements are in place with the AI vendors, and the company conducts penetration testing. Standard precautions, in other words, for a company handling sensitive health data.
Early traction metrics aren't public. The company launched in mid-February, which means it's been live for only weeks. Y Combinator backing and partnerships with established health data infrastructure players suggest serious intent, but intent and execution are different things.
The Real Test: Trust and Unit Economics
The central uncertainty isn't technical. The AI works, more or less. Wearables generate data. Electronic health records can be aggregated, though never as cleanly as anyone would like.
The uncertainty is twofold. First: Will people trust an AI to monitor their health continuously? There's a difference between using an app to check symptoms when you feel sick and letting software watch you all the time, ready to sound an alarm.
Second: Do the economics actually work? If human doctors are in the loop for every billable event, and if patients pay only when they escalate to a doctor, the model depends on a delicate balance. Too many escalations, and the doctors get overwhelmed. Too few, and there's no revenue.
Prana calls itself a "full-stack medical provider," though its legal disclosures paint a more modest picture: technology scaffolding around traditional licensed care. That's not necessarily a bad thing. Just a more honest framing.
The company is very young. The model is unproven. The market has been unkind to similar ideas before.
But the gap between annual checkups is real. And someone, eventually, might figure out how to fill it profitably.
Whether it's Prana remains to be seen.
