Raising Health Building the Marketplace for AI’s Most Valuable Asset
16 snips
Feb 17, 2026 Bobby Samuels, cofounder and CEO of Protégé, builds a marketplace linking proprietary real-world data to AI labs. He discusses sourcing and packaging multimodal longitudinal data, why real-world beats synthetic, the rise of eval datasets and benchmarks, legal and rights challenges, and expanding beyond healthcare into video, audio, motion, and biology.
AI Snips
Chapters
Transcript
Episode notes
Synthetic Data Is Complementary Not Panacea
- Synthetic and manufactured data have roles but don't replace real-world exhaust data for representativeness.
- Use synthetic data for prototyping, privacy-safe previews, or simple extrapolations, but train on real data for fidelity.
Stitching Multimodal Patient Journeys
- Protege built longitudinal, multimodal patient journeys by stitching EHRs, imaging, labs, and pathology across institutions.
- That capability differentiated them from traditional data players focused mainly on structured pharma-oriented datasets.
Design For Recurring Data Access
- Expect customers to request recurring access and refreshed datasets rather than one-off buys.
- Structure agreements for renewals and periodic refreshes so models stay current and useful.