All Projects

02AI Platform

Angel

Highlight a claim on any page and Angel digs up the primary document behind it. I built the LLM agent and the Chrome extension, plus the GCP setup underneath.

Role

Founding Engineer

Timeline

Jun 2026 - Present

Context

Stealth Startup - Helsinki

Core Stack

TypeScript, AI Agents, GCP

Angel screenshot 2
Angel screenshot 3
Angel screenshot 4

About the project

Angel is for journalists. Highlight a claim on any page, click Analyze, and it goes looking for the original document behind it. It doesn't tell you whether something is true. It tells you where it came from: a source, the excerpt that backs it up, and a verdict of source found, uncertain, or no verifiable record.

I'm the founding engineer. The agent runs on the Vercel AI SDK with required tool calling, so the model has to go and look things up instead of answering from memory. It searches the web, queries Statistics Finland directly, and opens candidate documents. Four output guards check the answer before anything reaches the user. A confident wrong source is worse than no answer at all.

I also own how it ships. Nx monorepo, Stripe billing, and GCP across three environments with GitHub Actions. Angel is also my MSc thesis at the University of Turku.

Highlights

Full Stack

TypeScriptVercel AI SDKNode.jsPostgreSQLNext.jsChrome MV3GCP Cloud RunFirebaseStripeNx Monorepo