Sept. 9, 2026— Atman Health has been selected by the Advanced Research Projects Agency for Health (ARPA-H) to develop an agentic artificial intelligence system for cardiovascular care, beginning with heart failure.
The award is made under ARPA‑H's ADVOCATE program (Agentic AI-Enabled Cardiovascular Care Transformation), a 39-month initiative to modernize how chronic cardiovascular disease is managed in the United States.
Approximately 7.7 million Americans live with heart failure, a condition treated with four classes of medication that must be introduced and increased gradually over months, each step checked against kidney function, other lab values and blood pressure levels. The regimen is among the most proven in medicine, yet in major patient registries only ~1 percent of eligible patients were on target doses of at least three of those foundational drug classes.
Access to cardiologists compounds the problem. A 2024 analysis in the Journal of the American College of Cardiology found that 1,454 of the 3,143 counties in the United States have no practicing cardiologist. Optimizing medications for millions of patients requires repeated clinical contact over many months, and an adequate clinical workforce does not exist.
"In heart failure, a single wrong adjustment can put a patient in the emergency department with elevated potassium, kidney injury or an arrhythmia," said Rahul Deo, MD, PhD, co-founder, chief product officer and chief medical officer of Atman Health. "That is why this work has remained manual. Every clinician has encountered an AI system that sounds correct but is often wrong, with no way to tell which. Before such a system can make these decisions on its own, a physician must be able to ask what was checked and why, and did it get the same answer every time."
Cardiovascular System
Atman's system separates two functions that most clinical AI systems combine. The large language model (LLM) interprets records and handles patient conversation by telephone and chat. However, clinical decisions are made outside the LLM, by a deterministic engine built on more than 9,000 explicit clinical criteria, 9,600 medication and indication pairings, 6,900 titration profiles and more than 100 symptom triage flows. It checks exclusion criteria before proposing any medication and records, for each recommendation, the logic that produced it, the evidence behind it, and the hazard classification under ISO 14971. A shadow agent trains and validates the system's behavior in real-world clinical settings.
"Encoding cardiology as an explicit rules-based system is slower and less fashionable than training a large language model, but we chose encoding deliberately," said Rahul Patel, co-founder and chief executive officer of Atman Health. "An LLM cannot be given permission to act autonomously on reasoning nobody can examine, and permission is the whole game here. Everything with a clinical consequence happens inside the encoded engine, in a structured form that a cardiologist, client, or regulator can review."
The regulatory path is defined: engagement with the U.S. Food and Drug Administration through the pre-submission process in the first year, an Investigational Device Exemption application at month 12 and a De Novo submission targeted within 24 months, preceded by a single-center shadow-mode study.
"Ten years ago, I said that you should be able to get better cardiac care in a village than at a leading academic medical center," said Vinod Khosla, founder of Khosla Ventures. "That becomes possible only when expertise stops being a scarce human resource and becomes something machines can reproduce. Reproducing it means the same evidence is applied the same way every time, and someone can check when it is not. Expertise will be nearly free. Atman is building the system that does that for cardiology and is building it to a standard the FDA can authorize."
September 09, 2026 
