Whether you need to architect multi-agent systems, optimize edge vision models, or audit LLM security, I'm open for advisory and contract engineering.
Autonomous agents and multi-agent workflows with LangGraph and CrewAI, tool use, function calling, and guardrails.
Outcome: A process that ran on manual effort now runs itself, with human-in-the-loop gates.
Retrieval-augmented generation over your documents with hybrid vector + BM25 search and cross-encoder re-ranking.
Outcome: Your team can ask questions of your own data and get cited answers.
React and Next.js frontends with FastAPI and Python backends, multi-provider LLM orchestration, and fallback chains.
Outcome: A polished product your customers can actually use.
YOLOv8 with TensorRT INT8 quantization-aware training and GStreamer on Jetson-class hardware, sub-100ms inference.
Outcome: Real-time defect and object detection running on the edge.
Architecture reviews, LLM cost optimization via smart routing (up to 65% reduction), and fairness auditing for EU AI Act and NIST compliance.
Outcome: A system that is cheaper to run and defensible to an auditor.