Open: 01_live-dashboards/oog-intake/index.html (self-contained, offline; demo brand)
Problem. Client onboarding needs a structured discovery questionnaire that's fast to edit and produces clean, machine-readable output.
What I built. A zero-framework intake tool where the entire questionnaire lives in data (schema.js) — 9 sections / ~25 fields with field types (text, single, multi, group, yesno) and conditional reveal logic. The engine renders the form, shows a progress bar, lets the user review before submit, then exports a typed JSON payload (and a mailto fallback). Change the questions by editing the schema — no code.
Engineering signal. Clean separation of data vs. logic, conditional UX, review-before-submit, and a deterministic JSON contract out — the same "schema in, structured out" discipline behind good agent/data pipelines.
Why it matters. Practical internal-tooling: fast to adapt per client, accessible, and produces structured data downstream systems (or an LLM) can consume directly.
This is a client intake and fact-finding tool, and the whole point is that the questionnaire is data, not code. Every question lives in a schema file with its type and conditional reveal logic, so changing the interview is a one-line edit. The user walks the sections, reviews, and it exports a clean typed JSON payload. That structured-output discipline is the same thing I'd want feeding a downstream system or an LLM, schema in, structured data out.
GG · Applied AI Engineering portfolio · synced to youdontneedmy.help design