I've built AI Revise V3 at LegalOn Technologies, reaching an approximate 60% win rate across most evaluation datasets, and designed its end-to-end evaluation system from metric computation through an LLM-based judge. I also shipped the Workflow Engine for Project Re: Invent, enabling natural-language authoring of complex workflows.
At Autify, I led machine learning technical direction and delivery across multiple projects. I deployed a Visual Regression API pipeline serving 14M+ requests on Knative and GKE, and trained an in-house transformer OCR model that achieved over 95% character recognition across 30K+ Japanese and English characters.
My work spans LLMs, agentic workflows, evaluation, computer vision, OCR, and production MLOps—from experimentation through deployment.
