Practical AI in Healthcare S1, E29 - Shashi Shankar, Co-founder & CEO, Novellia, Inc.
Mar 22, 2026
Shashi Shankar, co-founder and CEO of Novellia and former Genentech executive, built a patient-centered data solution after a family cancer journey. He explains how patient-authorized SMART on FHIR records create longitudinal real-world data. The conversation covers why past personal health record attempts failed, how AI spots clinical data errors, and the tradeoffs of trusting Big Tech with patient information.
AI Snips
Chapters
Transcript
Episode notes
Claims Data Is Broad And Slow
- Claims datasets are broad and lagged, limiting phenotype detail and near-term insights for pharma.
- Novellia aims to reduce lag and add granularity (tumor type, biomarkers, treatment sequences) missing from claims.
Single Site EMRs Miss Longitudinal Care
- Single-site EMR-derived RWD (e.g., Flatiron) gives depth at that site but misses multi-site patient journeys.
- Novellia complements these sources by stitching multi-site longitudinal records via patient authorization.
HIEs Are A Subset And Often Thin
- HIEs are limited: voluntary data contributions, opaque ingestion, and treatment-use access restrict research value.
- Shashi argues patient-sourced aggregation is closer to raw source data than HIEs.