About GenHealth.ai

GenHealth.ai is a startup building the next generation of healthcare solutions using generative AI. Our approach trains a transformer model on millions of patients’ encoded claims and clinical data to predict their futures. GenHealth’s model can be used as a digital twin or co-pilot supporting a broad set of use cases. Among them, health insurance plans can use GenHealth to better understand quality and risk, providers can find the next best action, pharma can simulate clinical trials, and life insurance can have more personalized actuarial analysis.

We are a team of entrepreneurs, engineers, security experts, and scientists that deeply understand the healthcare domain and we work deftly with the cutting edge of technology. We founded and built 1upHealth whose infrastructure for health plans manages Fast Healthcare Interoperability Resources (FHIR) data. We are taking lessons learned from managing over 50 million patients' data and are now developing novel generative AI algorithms utilizing that newly standardized data. In the near future, we believe most major health decisions will be supported by  an AI; We want to be that AI.

Current State of the Art

Major healthcare decisions are currently guided by evidence based medicine which require pointed questions and research for each individual hypothesis. Researchers and analysts build manual reports and classical statistical analytics for each question without being generalizable to other use cases, diseases, or populations. This grossly increases the cost of identifying solutions which are limited by human capital. New methods exist, and healthcare has yet to take advantage of recent tech advances, specifically those employing Transformers.

GenHealth’s Approach

Our product is the world’s first publicly available Generative Healthcare AI. It includes a foundation transformer model (new neural networks with excellent memory), an API, and user applications which can generate and predict future sequences of health events using historical events. Together they can be used to build a large variety of customized applications.

We’ve built a patent pending Generative Pre-Trained Transformer (GPT) AI which we call DOOG-E. DOOG-E is similar to models like ChatGPT because we also use Neural Network Transformers, but unlike ChatGPT, our model is not a Large Language Model (LLM) but rather a Large Medical Model (LMM). It uses medical events instead of words as its “language.” Our V1 model is trained on claims and clinical data from one million Medicare, Medicaid, and commercial patients. Our model uses 70 million parameters and 1 billion tokens. DOOG-E is trained on demographic and clinical data that are encoded into a series of temporal events. We’ve primarily used ICD, CPT, RxNorm and LOINC codes as representations of clinical events. We used those event sequences to train DOOG-E to predict future sequences.

Endgame

In line with our thesis that most major healthcare decisions will be supported by an AI, we aim to make GenHealth’s AI as intelligent, safe, and accessible as possible to the broader industry. We hope our AI at health plans, providers, pharma, app developers, and elsewhere in the industry will improve quality, reduce cost, and help people live longer healthier lives. Generative AI has a power unlike we’ve ever seen before. We aim to bring that power to the healthcare decisions that affect the daily lives of billions of people.

Leadership

GenHealth is led by the three founders: Ricky Sahu, Eric Marriott and Ethan Siegel. Accompanying the founders, Don Rucker, the former National Coordinator at the ONC, and Aneesh Chopra, the former CTO of the United States, are both advisors at the company.