At IronLabs AI, I build agent-optimization and model-routing services that autonomously improve agent code, benchmark changes in sandboxed environments, and route requests across LLMs. I also developed a Python SDK, plugin system, CLI, and GitHub-backed versioning for direct inference.
Previously at Mahyco, I built drone-image crop-yield prediction pipelines and reduced counting MAE by 25%. I’ve also built RAG products for chess coaching and company-document analysis, with published research in aerial surveillance and wearable-device stress prediction.
