At Crossing Hurdles, I analyze real-world software codebases and build benchmark tasks for Handshake AI, including terminal-based reasoning tasks and web-search benchmarks. For Snorkel AI, I debug CI and test-grading configurations and use mutation testing to improve evaluation accuracy.
At Canum.ai, I developed Python and FastAPI services to ingest and analyze security event data. I also integrated OpenAI GPT-3/4 APIs into pipelines that automated cybersecurity alert analysis.
I built ElevateCV AI, a FastAPI and React platform that turns resumes and job descriptions into structured JSON and scores their fit. Its grounding layer checks AI-written resumes against the source and flags unsupported claims.
I also built an evaluation harness with a hand-labeled dataset, a pytest suite, and CI through GitHub Actions to check skill matching and grounding.
