Most people think of CAR-T as a clinical/commercial story. It's also a genuinely interesting data engineering problem.
Think about what it takes to build a system that ingests clinical trial registries, patent filings, regulatory actions (FDA/EMA/NMPA), conference abstracts, earnings call transcripts, and real-world evidence — then turns all of that into a queryable, sourced, continuously-updated model of a therapeutic landscape.
That's the architecture behind modern pharma competitive intelligence platforms, and it's a good example of RAG + entity extraction + semantic modeling applied to a genuinely messy, high-stakes domain.
Wrote up how these systems work (and why static BI dashboards can't keep pace with a space like CAR-T) here: https://medium.com/@seosperreach/ai-car-t-intelligence-platform-2026-8d89f42734ef
Top comments (0)