07 · Agent ops and production

What it takes to keep an AI app working for real users: seeing inside every request, controlling cost and latency, surviving load and outages, and shipping changes safely.

You'll be able to: trace an agent end to end with OpenTelemetry, estimate and cut cost per task, design for retries, fallbacks and rate limits, load test, and deploy with eval gates and rollback.

#	Note	Level	Time
1	Observability and tracing	Intermediate	12 min
2	Cost and latency	Beginner	12 min
3	Reliability and scaling	Intermediate	12 min
4	Deploying an AI app	Beginner	10 min

Project: P5 · Production capstone takes the P4 agent to production: tracing, cost metrics, load testing, Docker, a threat model and a design doc.

Next: 08 · AI system design