At Fossible Works Inc., I built HydraAI, an autonomous agent embedded in ERPNext that lets business users customize their ERP in natural language. I developed it full-stack, from its LangGraph agent server to its React client and multi-tenant SaaS.
I also built an Inspect AI evaluation framework with 59 end-to-end tasks, outcome-based grading, and hallucinated-completion detection. It supports model selection and regression testing before each release.
At Abundant AI, I authored containerized RL benchmark tasks based on real bugs in open-source Python libraries and designed original tasks in distributed systems and differential privacy. As a Google Summer of Code contributor to RADIS, I added a three-temperature electronic non-equilibrium model for molecules, released in radis 0.17.1.

