I quit my job to build a 24/7 AI infra engineer

Until recently, I worked for a fintech startup. I enjoyed most of my job, except for the on-call rotations. After 5 years of grueling on-call toil on the fraud team I decided enough was enough.

Despite there being hundreds of observability and testing tools on the market, we still spent countless hours sifting through logs, graphs, and dashboards to root-cause incidents and write reports. Testing didn't do enough to catch bugs before they hit prod, and when they did raise alerts in staging we'd end up pushing to prod anyway. Poor codebase hygiene, stale feature flags, unactionable errors all meant our engineers didn't trust our CI/CD pipelines as much as they should have.

I've had an LLM-based solution cooking in the back of my head for some time, and last month I quit my job to work on it.

Foam is shaping up to be a 24/7 AI infra engineer. The end-goal is to build a tool that, given a trigger (e.g. code deployment), detects, records, and resolves issues across the stack autonomously. It plugs into your CI/CD pipeline, codebase, and observability tools and continuously monitors your dashboards.

I'm starting small with rollbar, grafana, and github integrations. What issues do you run into at your jobs? Let me know what you think of the idea and if you see any issues.

Interesting!