A guided video lesson
Agentic SDLC at Uber: From Platform Building Blocks to a Managed Software Factory
A source-grounded lesson on Uber's six-layer agent platform and the end-to-end loop it supports: research, design, coding, early validation, CI, review, and maintenance. The central idea is that useful agentic engineering depends less on a lone coding model than on controlled access, relevant context, prepared execution environments, reusable skills, staged evidence, and feedback loops.
Chapter pages
Choose where to begin
- 01The Software-Factory Thesis and Operating Baseline0:14 - 1:48
- 02Model Gateway: Controlled Model Access at Scale1:48 - 3:48
- 03MCP Gateway and Token-Efficient Tool Access3:48 - 5:43
- 04Pre-Provisioned Agent Execution Environments5:43 - 6:35
- 05DevPods Subcapability: A Unified Multi-Repository Workspace6:35 - 7:12
- 06Managed Skills and Their Improvement Lifecycle7:12 - 8:39
- 07Context Graphs for Faster, More Predictable Retrieval8:39 - 10:21
- 08Packaging the Platform in Cortana10:21 - 11:34
- 09From an Idea to Business Research11:34 - 12:32
- 10Requirements, Design, and Rapid Prototyping12:32 - 13:18
- 11Building with Minion Before CI13:18 - 14:15
- 12Densifying Validation into the Inner Loop14:15 - 14:59
- 13Self-Healing CI and Evidence-Bearing Code Review14:59 - 15:54
- 14Managed Maintenance and Feedback Loops15:54 - 17:16
- 15Scale Bottlenecks and the Build-Versus-Should-Build Decision17:16 - 18:22