Stephen Humbert

I’m a founder-operator building production AI systems that actually run in manufacturing and small-business operations. I design and ship evaluation-first LLM systems, then deploy the local inference infrastructure needed to replace “API-only” workflows with on-prem reliability.

As Founder & CEO of Applied Stoicism Labs, I operate a 3D-printing manufacturing startup with a live paying customer and lead the end-to-end build of an internal manufacturing OS. I’m the architect of CWF—spanning a React/Vite frontend, Fastify backend, Python CAD service, and a Supabase data layer—so our workflows are enforceable, observable, and fast.

I also build the AI decision layer: I designed and ran a custom LLM evaluation harness across multiple task categories using a rubric and LLM-as-judge (with human override flags). That benchmarking directly informed which local models could replace API calls per task, and I deployed an on-prem compute node running Ollama with a full Docker-based tool stack.

Beyond evaluation and inference, I ship AI-native products end-to-end—like Open Brain, a persistent semantic-memory system for Claude sessions using Supabase pgvector and MCP-exposed Edge Functions, and a production PWA that streams outputs via the Anthropic API. My ethos is simple: build systems that are measurable, constrained, and maintainable—and make them useful for real operations.