I've built AI-powered products and evaluation systems that make agent and LLM outputs more reliable in production.
At Handshake AI, I engineered 87 containerized AI-agent evaluation tasks across 10 domains. I build reproducible environments, hidden tests, correctness oracles, and automated verifiers for reliability, recovery, concurrency, and edge cases.
At Spottr, I co-founded an AI-first preventive health platform that turns biomarker, lifestyle, workout, and supplement data into personalized health protocols. I designed multi-stage LLM workflows, structured prompts, JSON schemas, validation logic, and guardrails for explainable recommendations.
At LinkedIn, I improved AI-powered Sales Navigator features, evaluated generative-model migrations, and resolved messaging authorization vulnerabilities. I also built a self-service Kafka debugging portal and supported streaming reliability through monitoring, redundancy checks, and incident response.

