At Hospital I Tingo María – EsSalud, I designed and implemented a RAG-based clinical-request auditor with an LLM agent that flags issues in clinical histories and diagnostic coding before records are submitted. In evaluation on 50 labeled real cases, it reached 74–80% accuracy, with no invented data and all 11 adversarial cases rejected.
I also built data pipelines to extract, anonymize, normalize, and import clinical records from ESSI and REFCON, institutional systems without public APIs. My Python analysis of 15 months of historical data supported a proposal to install a CT scanner and surfaced a 399% gap between demand and capacity in General Medicine.
As an independent software developer, I delivered middleware connecting a clinic’s HL7 infrastructure to an AMIS-850 medication dispenser that exposes only a proprietary SDK, and developed a pharmacy point-of-sale and inventory system with DIGEMID registration and FEFO lot tracking. My projects include clinical referral services and a pediatric clinic management platform, using Python, FastAPI, Django, and PostgreSQL.
