EpiCast

First Place, Google Health AI MedGemma Impact Challenge, 2026. USD 30,000 grand prize, selected first from over 850 teams and 6,500 entrants worldwide.

Multi-Modal AI Epidemiological Surveillance System

Built a mobile-first disease surveillance platform for West Africa (ECOWAS) combining fine-tuned medical LLMs with audio AI for syndromic monitoring and cough-based respiratory illness detection.

Fine-tuned and quantized MedGemma 4B to GGUF format for on-device inference, enabling a complete offline AI stack on mobile devices with sub-second response times and zero cloud dependency.

Integrated MedSigLIP for clinical image triage, HeAR audio models for cough classification, and a multilingual NLP pipeline trained on 9,700+ corpus-grounded examples across French, Hausa, and Yoruba.

Designed a 4-tab mobile interface (Dashboard, Intake, Alerts, Reports) with real-time syndromic heatmaps and structured WHO IDSR situation reports generated by MedGemma 27B.

Tech Stack: MedGemma 4B/27B, MedSigLIP, Google HeAR, FastAPI, RunPod Serverless, Python