Agents & Tooling 路 Retrieval & Ranking 路 Multimodal Robotics
I build applied AI systems at the intersection of agent engineering, information retrieval, and real-world interaction. With a background in mathematics, I care about the details that make a prototype dependable: explicit permissions, traceable data, reproducible evaluation, and human oversight.
My work spans enterprise knowledge workflows, candidate retrieval and ranking, and multimodal sensing. I enjoy connecting models to useful tools鈥攁nd checking that the resulting system does what it claims.
| Area | What I work on |
|---|---|
| Agent systems | MCP tools, structured outputs, human approval flows, permission-aware execution, and idempotent writes. |
| Retrieval & evaluation | Hybrid lexical and semantic retrieval, two-stage ranking, held-out evaluation, and data leakage prevention. |
| Multimodal robotics | Depth cameras, IMUs, voice interaction, sensor pipelines, and separating model suggestions from hardware execution. |
| Applied ML | Low-cost spectral sensing, sample-level validation, matched ablations, and end-to-end experimental demos. |
Languages 路 Python 路 SQL 路 Rust (working knowledge)
AI systems 路 LangGraph 路 RAG 路 MCP 路 Tool Calling 路 Human-in-the-loop
Infrastructure 路 PostgreSQL 路 pgvector 路 Docker 路 Git
Evaluation 路 Recall@K 路 NDCG 路 Precision 路 Latency 路 Ablation studies
Selected experience is described at a technical level; client identities, private datasets, and internal implementation details are omitted.