Projects

Leading the development of an AI platform designed to detect, map, and respond to AI-amplified threats against institutional trust in real-time. A structural response to the commercialized ecosystem of disinformation — rooted in years of health equity research and crisis informatics. Trust in critical institutions — health systems, defense agencies, and emergency responders — is under systematic attack; generative AI has removed cost and skill barriers to industrialized misinformation campaigns. Three pillars: SEE (real-time multimodal detection — MedContext deployed at 91.4% accuracy for medical misinformation); MAP (graph neural network analysis of narrative spread, in development); ACT (operational response including pre-bunking and counter-messaging, in development).

Interdisciplinary research dedicated to equity, trust, and resilience in crisis response. In times of crisis — from pandemics to climate emergencies — misinformation undermines trust in science and exacerbates health inequities. This work brings together frontline clinical expertise, cutting-edge AI, and global health leadership to transform evidence into action. Research pillars: Trust & Crisis Intelligence (AI tools to detect and counter misinformation); Resilient & Equitable Health Systems; Equity-Driven Interventions. A unique pipeline from evidence to actionable AI systems deployed at scale.

Private consulting, strategic guidance, and investment in global health, technology, and AI readiness.

Multimodal detection across open and social sources, achieving 91.4% accuracy in detecting medical misinformation — particularly authentic images paired with false claims. Evaluates contextual authenticity rather than content alone. Submitted to the Kaggle MedGemma Impact Challenge. Powers the SEE (real-time detection) pillar of the trust defense platform.

A focused desktop tool for keeping track of papers, submissions, and revisions. Each manuscript becomes a complete record — authors, versions, submission state, and journal requirements — rather than scattered notes and PDFs. Local-first, no cloud or seat licensing. Built for academics who, like me, were tired of losing track of which version of which paper went to which journal.