Selected work
These are not exhaustive case studies. They are concise summaries of the kinds of AI systems and engineering problems I spend most of my time on.
2023 to present
Agent-assisted review workflow integrated into CI environments to reduce turnaround while keeping human judgment in the approval path.
The emphasis was operational usefulness: clearer review loops, faster iteration, and less noise for the engineers receiving suggestions.
LLM / agentic / devops
2022 to present
Production computer vision for inspection lines, built with attention to latency, inference constraints, and the practical realities of industrial use.
The work spans model behavior, serving paths, and the surrounding engineering needed to make deployment viable over time.
CV / C++ / edge
2023
Retrieval over dense engineering documentation for teams that needed direct answers instead of another search interface.
The goal was not broad consumer chat behavior. It was narrower, higher-signal retrieval aligned to practical internal usage.
RAG / LLM / backend
2021 to present
Internal tooling for managing, training, and monitoring computer vision models used by non-specialist teams.
The value came from reducing operational friction and making model workflows more legible to the people actually using them.
full-stack / UX / ML ops