Max Laurie Hutchinson
12+ years building production software, regulated data platforms and distributed systems. I now apply that foundation to agentic AI, evaluation, retrieval and governed automation.
Current work
Public personal engineering
Agent evaluation and orchestration
A Python evaluation runtime for Agent Skills. A dependency graph schedules evaluators, joins their evidence and gates whether to accept, revise or escalate.
Working prototype
Controls around model output
A Python CLI for policy retrieval, deterministic checks and optional model analysis. Unsupported citations fail closed; approval stays with a human.
Internal product
Product knowledge with a deliberate access boundary
Python retrieval from a real Seller account and a separate Semantic Kernel RAG application, using vector embeddings in Cosmos DB. Material changes stayed manual.
How I work
Keep systems small. Add complexity only when evidence earns it.
Make failures visible. Design for observability, recovery and replay.
Give models useful work and clear limits. Keep authority over consequential actions explicit.
Writing
Medium · Featured
An Outreach Skill pilot made me question what we give away when we tell an AI reviewer what to look for.
Medium ·
A practical account of separating automated memory maintenance from human approval of changes. Published on Medium in February 2026; tooling and pricing reflect that date.
On this site ·
Separating signal, policy, and execution gives you a system you can replay, audit, and evolve without pretending the model is perfect.
I work across architecture, implementation and technical leadership. Tell me about the problem, the team and what needs to change.