Learning roadmap· 100% match
AI Ops
Run AI in the real world: keeping it fast, affordable, reliable, and watched — the things nobody warns you about.
5 stages~10 h
That leaves 5 stages, ~10 hours of focused work. You can clear it in any order, but the stages compound — try them top-to-bottom your first time.
- Conceptsedge-runtimestreaming-ssecold-startregion-pinningLab task
Deploy the same chat to edge AND origin. Measure cold start + p50/p99 latency.
CheckpointYou can name your deployment’s p50, p99, and cold-start cost from memory.
What you'll be able to do
- Run AI in production with a credible 99.5% availability target.
- Have a fallback chain so a single provider outage doesn’t take you down.
- Get paged for the right things, not noise.
Or try a related path