What This Is
An agent-based system that ingests a patient's health signals (genomic data, lab results, EHR records, continuous wearables, lifestyle data, medication history) and generates personalized clinical insights that are specific to that individual's biology, environment, and risk profile. Not a diagnostic tool. Not a replacement for clinicians. A contextualized decision support layer that lives between a patient's data and their doctor's judgment.
The system works by enrolling a user, collecting their data sources (with appropriate consent and integration), and running that data through a Claude-based agent that's been trained on clinical literature, guideline summaries, and population studies. The agent outputs a report that highlights what matters to this specific person: which preventive screenings are urgent for their genetic risk, which medication interactions matter given their genome, how their wearables compare to their peer group, and what lifestyle changes would move the needle on their specific risk profile.
The core thesis: personalization only works at the data intersection. A single genomic variant means little without lab values. Lab values mean little without family history. Family history means little without your actual behavior data. The software is the connective tissue.
Why It's Interesting
Three hard facts converge. First, genomic sequencing cost has dropped from $3 billion to $100 per person in twenty years and continues falling. Second, we have 15+ years of evidence that genome-informed care reduces disease risk when paired with lifestyle or medication changes. Third, we have continuous wearables (Apple Watch, Oura, glucose monitors) capturing daily signal in hundreds of millions of people, yet almost none of this data is integrated with traditional medical care.
The market appears addressable: 50 million Americans have taken a direct-to-consumer DNA test. Roughly 200 million have wearables. Chronic disease costs the US $4 trillion annually and kills 7 out of 10 Americans. If personalized medicine prevents 5 percent of preventable deaths or hospitalizations, the value is real.
The timing is right because the infrastructure exists now. EHR APIs are open. Genomics companies will share data. Wearables have APIs. And Claude's reasoning ability makes it possible to write a single agent that integrates across modalities instead of building ten separate rule engines.
Why A Landing Page Would Fail
A landing page sells the dream: "Know your health, live longer." Reality is friction at every step. First, regulatory. The FDA has not yet established a clear pathway for "AI clinical decision support" and the conversation is still happening. Deploying something that looks like medical advice without that clarity gets you a warning letter. Second, integration. Connecting to EHRs requires HL7 compliance and often partnership with health systems. Connecting to wearables requires constant API maintenance. Most of your users won't connect their data because the onboarding is a credential form followed by "checking with Epic" for 48 hours. Third, liability. Even with disclaimers, the moment someone follows your advice and has a bad outcome, you are the software company they sue. Fourth, trust. Doctors don't refer patients to software they don't know. Patients don't pay for software that doesn't come from their doctor. You need institutional distribution, not consumer word-of-mouth.
The landing page tells a cohesive story. The business is all friction and risk concentration.
The Realistic Shape
Start narrowly inside health systems, not DTC. Partner with 1-2 health systems that have mature EHR ecosystems and willing cardiologists or oncologists who want better preventive risk stratification for their patient panels. Integrate their EHR API, add optional wearables and genomics, deliver reports to the clinician, not the patient. The clinician decides what to do with the insight.
Team: 2-3 engineers (integrations, agent tuning, data validation), 1 clinical advisor (MD or genetic counselor), 1 regulatory consultant, and Ankit as product. You need the clinician to survive the conversation with health system medical directors. You need the regulatory person to navigate slowly.
Capital: $2-3M seed. Runway is 24 months minimum because pilot timelines inside health systems run 12-18 months before you get any traction data. Burn rate is low because you're not running marketing, and the pilot is unpaid (you're validating).
Six-month milestones: (1) Partnership letter in hand. (2) Basic EHR integration working with one data source. (3) Agent pipeline integrated with at least genomics or wearables. (4) First 50 patients in pilot with clinician feedback. (5) Regulatory status mapped. (6) Second health system in conversation phase.
Honest 12-Month Case
Best case: first health system loves it, shows clinically meaningful shifts in screening compliance or risk detection, pilots expand to two more systems, you land a Series A on the strength of pilots and institutional traction. Revenue is still zero but your risk is de-risked because health systems have said yes.
Mid case: first pilot shows uptake but no clear clinical signal in 12 months, but doesn't fail. You raise Series A at a lower valuation, focus on refining the clinical protocol, expand to more patients in the same system, and start thinking about payer partnerships (because insurance will pay for preventive care if it reduces expensive acute events). Revenue stays zero but the category is validated.
Bad case: integration hell. The health system IT team delays 6 months. The clinician loses interest. You're burning cash and have no evidence that doctors want this product. You either pivot to DTC (which means rebuilding with regulatory and consumer friction), or you shut down.
Kill criteria: By month 12, if you don't have a clear pathway to either regulatory approval or a health system paying pilot, the bet is unlikely to hit.
Five Questions Before Committing
1. Which health system is signed? Do you have a named champion physician and timeline?
2. What is your clinical differentiation over existing risk engines from Tempus, Color, or Invitae? Why will a cardiologist use your report instead of their existing tool?
3. If you get 90 percent EHR connectivity and 50 percent wearables connectivity, is the product still valuable, or does it require all three data sources?
4. What is the first revenue model: health system pays, insurance pays, patient pays, or non-negotiable free pilots until traction?
5. Who on the team has FDA interactions experience or clinical integration experience with a major EHR? This project dies without it.