At OpenNG, I manage and maintain the backend codebase for a healthcare application supporting patient registration, diagnosis logging, and hospital payment workflows. I also built Sirius, an agentic learning backend that turns natural-language goals into validated curricula, lessons, and practice.
In Sirius, specialized agents work through durable workflows while deterministic services handle validation, evaluation, mastery, and progression. My research projects include a MuZero agent with latent-state PUCT for ForexMind and Mamba-2 experiments for ArcRSM, alongside position-based chess knowledge retrieval in ChessMind.
