AI/ML engineer in San Francisco building retrieval pipelines, agentic systems, and vision-language applications. I focus on the engineering around the model: orchestration, evaluation, observability, and reproducible experiments.
LinkedIn | Open-source contribution
A cost profiler for voice-AI agents, connecting trace pricing, outcome adjudication, waste detection, and replay experiments in a Python workspace.
Includes synthetic workload re-pricing and controlled local TTS experiments, with measured and modeled results labeled separately.
A learning-roadmap generator with a FastAPI backend and React interface. It combines Qdrant dense/sparse retrieval, cross-encoder reranking, and bounded query reformulation with web fallback for weak matches.
The repository includes the retrieval implementation, test cases, and evaluation methodology.
Contributed LangGraphKernel to the open-source microsoft/agent-governance-toolkit repository: pre-compile node wrapping, policy hooks, and checkpoint authorization fingerprinting.
PR #2694 was merged and the adapter was included in release v4.0.0.
- Agent systems: Python, LangGraph, FastAPI, tracing, and policy-aware orchestration.
- Retrieval: Qdrant, hybrid search, reranking, and evaluation.
- Vision-language modeling: LLaVA, QLoRA, PEFT, and PyTorch.
Interested in applied AI/ML engineering and hands-on customer-facing implementation work. Connect with me on LinkedIn.