A guided video lesson

Agentic SDLC at Uber: A Managed Software Factory

How shared model, tool, runtime, skill, context, and assistant layers connect ideation, coding, validation, review, and maintenance while preserving guardrails and feedback loops.

Open the complete source

Chapter pages

Choose where to begin

  1. 01Why Uber frames agentic development as a software factory0:14 - 1:50
  2. 02A model gateway for privacy, policy, latency, and attribution1:50 - 3:48
  3. 03An MCP gateway that makes tools usable and affordable3:48 - 5:43
  4. 04Pre-provisioned environments for long-running, cross-repository agents5:43 - 7:12
  5. 05A managed lifecycle for reusable agent skills7:12 - 8:39
  6. 06A context graph for finding software and business context8:39 - 10:21
  7. 07Packaging the platform in Cortana and everyday surfaces10:21 - 11:34
  8. 08From a Slack idea to research, requirements, and a prototype11:34 - 13:18
  9. 09Minion turns the prototype plan into a draft change13:18 - 14:18
  10. 10Move validation inward, then make the outer loop self-healing and reviewable14:18 - 15:54
  11. 11Maintenance as a managed, learning loop15:54 - 17:16
  12. 12The next bottlenecks: monitoring, capacity, experiments, and judgment17:16 - 18:22