Agentic AI and LLMs
I design agentic harnesses and integrate LLMs into business software, with the same rigor as a critical system.
AI is nothing new to me: expert systems for fault diagnosis and industrial planning early in my career, machine learning and deep learning at Systel, LLM agents today.
- Agentic harnesses: tool loop, context management, error recovery
- Agent frameworks (LangChain, LangGraph, Claude Agent SDK) and multi-agent orchestration
- MCP servers to connect agents to your systems
- RAG and embeddings over your business documents
- Interchangeable LLM providers, with no vendor lock-in
- Guardrails: human approval, permissions, audit log, cost control
Rust, TypeScript, Python, LangChain, LangGraph, Claude Agent SDK, MCP