Navia

Clinicians, researchers and pharma sponsors were losing whole days moving between ClinicalTrials.gov, Orphanet, PubMed, EuropePMC, Open Targets, EMA EPAR, WHO ICTRP, CTIS and the European Reference Networks. Every source names diseases differently, so nothing lined up. Navia was built to end that.

I built the entire stack from an empty repository. FastAPI and PostgreSQL hold the data, Celery and Redis run ingestion, and a Next.js front end presents it. The ingestion framework is idempotent, so a rerun never duplicates a record, and every trial condition is mapped back to a canonical disease so counts can be trusted.

On top of that sits a snapshot pipeline that computes more than twenty five derived metrics per disease, a comparison engine for up to five diseases at once with PDF export, and an AI summary layer with circuit breaker protection so a slow model never takes the page down with it.