Roadmap
The AI engineer roadmap
Start with the short path, use the complete curriculum as a reference, then choose one specialization. Every path builds the same real project and measures quality, latency and cost together.
0/100
core essentials done
core depth 0/201
everything 0/332
The shared core
Build one project through four phases
Follow these phases in order. Learn both retrieval and agent workflows at working depth, make the same project measurable and safe, then specialize only after you can ship it.
Follow in order
Your path from foundation to production
Each module below is the same canonical module in the complete curriculum. Open it to study topics, resources and its completion project.
- 01
Phase 1
Learn the foundations
Understand the software boundary, the model boundary and the vocabulary used everywhere else.
0/290%Phase outcome: A typed LLM API with structured output, automated checks and a measured quality, latency and cost baseline. - 02
Phase 2
Build useful AI systems
Give the application reliable knowledge and controlled ways to take action.
0/200%Phase outcome: One cited retrieval flow and one checkpointed tool workflow with retry, approval and stop conditions. - 03
Phase 3
Make it trustworthy
Prove that the system works, stays inside its budget and respects its trust boundaries.
0/320%Phase outcome: A regression suite, end-to-end traces, per-task budgets and a tested threat model for the same project. - 04
Phase 4
Design and ship
Turn the working feature into a defensible system and evidence that you can operate it.
0/190%Phase outcome: A deployed capstone with an SLO, rollback path, design document, public demo and honest postmortem.
The fork
Then pick exactly one of 8 paths
The core above is shared by every role. Each path below adds depth on top of it — you need one, not all 8.