Where I’m focused this year

Amoebanator — clinical ML safety

A 9-class differential diagnosis system for severe central nervous system infections, with primary amoebic meningoencephalitis as the highest-stakes target class. PyTorch on RigoBERTa Clinical, 5-model deep ensemble, dual-gate out-of-distribution detection (Mahalanobis plus logit-energy), Mondrian per-class conformal prediction, and a hardcoded safety rule for freshwater exposure. medRxiv preprint targeted for May 28, 2026.

Kallpa — bilingual primary care decision support

A bilingual Spanish/English clinical decision support engine for primary care, deployed at kallpahealthcare.com. V1.0 is scoped to 200 conditions with a target ship of October 2026. The longer-term roadmap covers Peru, Latin America, Europe, and the United States.

What I’m looking for

• Co-authors and adjudicators on Amoebanator: clinicians with severe CNS infection experience, in Spanish or English.
• Pilot sites for Kallpa: primary care clinics in Peru, with priority on bilingual or rural settings.
• Conversations with PIs in computer science, biomedical informatics, clinical ML safety, conformal prediction, multilingual clinical NLP, or bilingual healthcare AI for underserved populations.