Public repos. Real code, not theory.

A scoring rig that returns a 0–100 number and the reasoning behind it.

Solves AI scores things 0–100 but won't say why, so you can't trust the number or defend it to a stakeholder.

Turns screen recordings and call audio into structured, queryable feedback.

Solves Hours of recordings and bug reports never become anything you can search, categorize or prioritize against.

Routes a task to a model, then falls back to rules or another model when it fails.

Solves A single model call is a single point of failure — when it hallucinates or times out, the whole run dies.

Normalizes and enriches messy records into linked Notion databases with integrity checks.

Solves Source data lives in scattered spreadsheets and scraped HTML, and syncing it into Notion by hand never stays consistent.

Turns meeting transcripts into deduplicated Notion tasks on a schedule.

Solves Action items from dozens of weekly meetings get lost in transcripts or copy-pasted by hand into the wrong tracker.

Versioned, reviewable prompt and spec templates that live next to the code.

Solves Your best prompts live in chat history, so you rewrite them from scratch every time and can't review or hand them off.