ai-platform/roadmap
All pages

Start here

  1. How to use this roadmap
  2. Two kinds of role
  3. Read code, write it rarely
  4. Target profile
  5. Eight principles
  6. Work on a real product

Phases

  1. 01Direct coding agents and read their code
  2. 02Learn how models behave and what they cost
  3. 03Build retrieval that finds the right evidence
  4. 04Build tool servers with MCP
  5. 05Design agent workflows that survive failure
  6. 06Run agent workloads on AWS
  7. 07Trace every step an agent takes
  8. 08Build evals that catch regressions
  9. 09Run agents reliably in production
  10. 10Secure agents and the code they write
  11. 11Build an internal AI platform
  12. 12Get teams to adopt the platform
  13. 13Build Forge, the capstone platform
  14. Elective: Kubernetes

After the phases

  1. Staff skills
  2. 12-month plan
  3. A typical week
  4. Show your work
  5. Tools to know
  6. What to skip
  7. Final checklist

After the phases

A typical week

You learn fastest in short cycles of building, measuring, and adjusting. Do not spend a month on courses before you touch the system. Every week should leave the system more capable or more reliable.

Pick one capability in the first session

Pick one narrow result, such as "add trace propagation through SQS", "measure retrieval recall at 5", or "require approval before write tools". Before any agent starts, write the expected behaviour and the acceptance tests as a short spec.

Direct and check in the build sessions

Run agents in parallel, each in its own worktree, and let your spec and context files guide them. For anything architectural, ask for options and trade-offs before the agent writes code. Then check the result in this order: tests first, the riskiest parts of the diff second, and a real run third. You are training the skill of deciding what is safe to merge.

Add no features in the reliability session

Spend one session a week without adding features. Break the system, read traces, improve tests and evals, add a missing metric, simplify an interface, or reduce cost. This session keeps you from building an impressive demo that fails in production.

Record the evidence in the last session

Update the README or the architecture diagram, write a short decision record, record before and after metrics, and log any failure that should become an eval. If you had to tell an agent the same thing twice, add it to a context file or a skill.