Skip to content
View shawnpetros's full-sized avatar

Block or report shawnpetros

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shawnpetros/README.md

👋 Hey, I'm Shawn.

I'm the dude who took apart toasters as a kid to find out why AND how they worked.

Fifteen years in tech turned that into React, TypeScript, AI pipelines, and an unhealthy number of repos. Then the industry buried the fun part under DSA interviews and "quality with velocity" metrics... so I went and got it back.

Now I build things that light up when I tell them to. And it feels like the first time I put a hover state on a button with jQuery.

What I'm Building

  • open-brain — Semantic memory layer for AI agents. Postgres + pgvector + MCP. Capture a thought, it gets embedded, classified, and filed. Ask a question in natural language, it understands what you meant. 7 tools, one brain. TypeScript + Python.

  • mediascribe — Extract structured insights from audio and video recordings. On PyPI and Homebrew. Python.

  • WindWatts — Open-source wind energy assessment tool for the National Lab of the Rockies. Mobile UI, public API docs, shareable deep-link architecture. React, TypeScript, FastAPI. 25 PRs merged.

  • claude-code-clone — Built a minimal clone of Claude Code from scratch. 42 features, zero human-written code, three agent sessions. The agentic loop is just send, detect, execute, feed back, repeat. That's it.

  • salazar — AI agent framework. The one I actually use in production.

What I Actually Care About

Vibe coding with guardrails. Spec-driven development where the AI handles implementation and the human architects the output. Agentic workflows, MCP servers, CLI tools designed for LLM consumption... and putting the art back into a profession that forgot it was one.

Writing a field guide about it. Working title: Kindling.

Petros Industries

Staff engineer turned independent. I build AI tools and automation for small businesses — the operations layer they can't afford to hire for. petrosindustries.com


📍 Los Angeles · shawnpetros.com · LinkedIn

Pinned Loading

  1. salazar salazar Public

    A harness for invoking long running agent loops.

    TypeScript 17 1

  2. claude-code-clone claude-code-clone Public

    just trying to learn how this all works...

    TypeScript

  3. NatLabRockies/windwatts NatLabRockies/windwatts Public

    TypeScript 4 10

  4. open-brain open-brain Public

    Semantic memory layer for AI agents. Postgres + pgvector + MCP. Capture, relate, retrieve.

    TypeScript 1

  5. mediascribe mediascribe Public

    Python