Fifteen-plus years designing certification and enablement programs at scale across Meta, Google, and Apple — and, on the side, shipping AI-native products solo on the Claude API. The work below is that second half: things I've architected and built end to end, not just taught.
Enable
Owned a learning platform with 27,000+ active users. Led teams up to 10. Programs shipped in 10+ languages.
Build
Ships production AI tools solo. Sigil — an adaptive certification platform — built end to end in a single weekend on the Claude API.
Selected Work
04
Certification · SaaS
Sigil
Adaptive certification that builds and grades itself.
An adaptive certification platform that generates assessments and adjusts them to each candidate in real time — built solo over a single weekend on the Claude API, from data model to grading logic to interface.
What it provesI can architect and ship the exact category of product I'd be enabling — closing the usual gap between someone who teaches a platform and someone who could have built it.
You play the gateway itself — checking each request's key against its badge, stamping honest status codes, and living with the consequences of every call. An in-experience coach grades free-text answers live on the Claude API, with a scripted fallback and xAPI throughout; three chapters, built solo.
What it provesBoth halves of my work in a single build — rigorous, measurable learning design and hands-on AI product engineering — the exact intersection I operate in.
An AI-coached skill-development platform (2026) that applies conversational AI to structured, ongoing practice — the learning domain I've specialized in for over a decade, rebuilt around what AI coaching now makes possible.
What it provesSustained product development, not a one-off — and it puts my enablement expertise and my building in the same place.
Turns tabletop sessions into finished comic pages, automatically.
A multi-stage pipeline that takes a raw D&D session recording and produces illustrated comic pages: Whisper for transcription, an LLM for story generation, a custom Python LayoutDirector for composition across seven layout templates, and ComfyUI for image generation — running locally, end to end.
What it provesI can design and orchestrate a real pipeline across the modern AI stack — transcription, generation, layout logic, image models — not just write prompts. Systems thinking on a genuinely hard multi-model problem.