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

Start here

What a finished AI platform engineer can do

An AI platform engineer combines backend, distributed-systems, site reliability engineering (SRE), security, developer-experience, and AI skills, and does most of the building through coding agents. The table lists each capability, what good looks like, and the proof you can show.

CapabilityWhat good looks likeProof you can show
Working with coding agentsRuns several coding agents at once, each isolated, with specs and review steps. Configures agent tools for a team: context files, skills, hooks, Model Context Protocol (MCP) servers, and permissions.A documented agent setup that other engineers use, with numbers for throughput and defects.
Code literacyReads Python and TypeScript fluently, and C# where needed. Finds concurrency, error-handling, security, and test gaps in diffs that agents wrote.Review notes on real pull requests, a planted-bug hit rate, and incidents debugged to the root cause.
Distributed systemsDesigns retries, queues, idempotency, backpressure, state, and recovery.Agent workflows that survive partial failure and that you can replay safely.
Cloud platformKnows AWS compute, messaging, storage, identity and access management (IAM), networking, and infrastructure as code.Environments that Terraform or the AWS Cloud Development Kit (CDK) creates the same way every time, deployed through continuous integration (CI).
LLM systemsKnows how large language models (LLMs) behave and what drives cost. Understands context, tools, and retrieval.Model routing and context choices backed by measurements.
Agent workflowsUses state machines, routers, supervisors, and human approval where each one fits.Multi-step workflows that survive restarts.
Reliability and evalsDefines service level objectives (SLOs), traces behaviour, and catches regressions before release.Eval suites, dashboards, alerts, and incident runbooks.
SecurityTreats model output, retrieved content, and agent-written code as untrusted input.Tools with least privilege, sandboxes, approvals, and audit logs.
Product judgementDecides what to build and when output is good enough for users. Talks to customers directly.A product with real users, usage numbers, and a record of changes made from their feedback.
Developer platformBuilds defaults that developers and their agents choose because the defaults save them work.SDKs, templates, shared skills and MCP servers, docs, and adoption numbers.
Staff leadershipMakes architecture decisions, writes requests for comments (RFCs), and influences teams outside their own.Adopted standards, design reviews, mentoring, and incidents led.