Work With Me
I build LLM systems that are measured, not guessed: retrieval pipelines, agents with verification layers, and evaluation harnesses that tell you whether the system actually works.
What I can do for you
- RAG systems: ingestion, hybrid retrieval, reranking, and answer quality evaluation
- LLM agents: tool use, multi-agent pipelines, and verifier layers for high-risk actions
- Cost and latency optimization for existing LLM apps
- Evaluation: test sets, automated scoring, and regression checks before release
Selected work
- Adaptive top-k RAG router: cut input tokens by 17% against a fixed top-k baseline with no extra LLM calls. View code
- Multi-agent planner with a verifier and inspector: LLM-planned robot tasks checked by a RAG-grounded safety layer in simulation. View code
- Safety-aware LLM task planner: runtime safety layer benchmarked against naive execution in PyBullet. View code
Background
Co-founder at Arbotrix (autonomous robotics), published research on safety in human-robot collaboration (IEEE Xplore), and several years of freelance AI/ML work for international clients.
How to start
Send a short note about your project, timeline and budget through the Contact page. I reply within 2 working days.
Comments