Andrew Fribush

I'm a software engineer in Austin. I run the AI practice at Abstract — speech, vision, and knowledge systems for clients who need them to hold up in production — and on my own time I turn US federal datasets into MCP servers and REST APIs, because the government publishes some of the best data in the world in some of the worst formats.

I like the unglamorous parts of the job: ingestion, compliance, the last ten percent that separates a demo from a product. If you have a problem shaped like that, I consult.

This site is a build too: it has a command palette (⌘K), a working terminal, an X-ray mode, and a resident wren who answers questions. The design is deliberately plain; the interesting parts are underneath. How it works →

fig. 1.1.2 — the return from Troy, drawn to men remaining.

fig of the month · aug 2026

source: Odyssey body counts · odysseyLosses.data.ts

Selected work

Dendrite
The corpus engine behind the fleet, one config turns a raw dataset into an MCP server, a REST API, and a static site.
Cairn
A zero-runtime continuity layer over git and files that stops long-horizon LLM work from forgetting or confabulating.
ALL-SAFE
An AI assessment OS for surgical training, scaffolding modules through agentic automation, combining peer, expert, and automated review, with AI feedback and video tuned for low-resource settings.
Credo Beauty
Multi-year work on the largest clean-beauty retailer in the US, Shopify Plus, deep NetSuite ERP and POS integration, and a mobile app taken from zero to over $5M a year.
Spartacus
A platform for collective-action campaigns powered by conditional commitments, pledges that only trigger once a threshold is met.

All projects →

Elsewhere

GitHub · LinkedIn · Twitter · hello@andrewfribush.com