AI engineering
I build the systems and operate them in production.
Shipping is the start. I run what I build: migrations, evaluations, cost, uptime, and the 2am text when something breaks. That operating experience is why the things I design keep working after the demo.
What that looks like
Production AI automations
Scheduled pipelines and agentic jobs that run unattended and fail loudly instead of silently.
Model migrations and evaluation
Rollback criteria defined before the switch, and evals that say whether it actually worked.
Full-stack builds
TypeScript, Python, Next.js, AWS, Playwright. The stack on the resume is the stack in production.
Operations
Alerting people do not learn to ignore, and runbooks a successor can follow without me.
Case studies
A fleet, not a feature
Hundreds of automated jobs at a large news publisher run through a platform I own, across 9 markets.
Read the case study →Collection at scale
A measurement platform collecting ~1.35M AI responses a month, unattended, built and operated solo.
Read the case study →
Have a system that needs building, or one that needs to stop breaking?
Email peter@peterbittner.com