At Microsoft, I lead the science and technical work for GitHub Copilot CLI’s auto-approve feature. I own its system architecture, classifier, evaluation, and telemetry, and serve as the technical DRI for launch criteria and rollout.
At Splunk / Cisco, I architected the Investigation Report skill for the Security AI Assistant, lifting its skill rating from 3.0 to 4.5 out of 5. I also designed dynamic query routing and served as tech lead for an agentic workflow automation engine.
My work spans applied AI and speech research. At Lightbird, I led speech and text features for a conversational AI product and fine-tuned Whisper for a 25% quality gain over the base model; at Columbia University, I research neural mechanisms of speech processing and teach graduate Generative AI.

