ROADMAP · 40+ hours of focused reading; less over months
AI Engineer · the full roadmap
You want the full arc, no shortcuts. Or you're not sure which specialized route fits — start here, branch when it's obvious.
Who this is for
Two shapes of reader end up here:
- The completionist. You want to know every corner before you move.
- The confused. You’re not sure which specialized roadmap fits — this one has everything, so you can branch when it becomes obvious which arc is yours.
What “the full roadmap” actually is
Eight tracks, sequenced. Each track has 3-6 lessons plus its own case files, interview cards, and problems. Read straight through if you’re patient; skip ahead by track if you already ship in that area.
When to leave this roadmap for a specialized one
- You’re prepping for an interview →
/roadmap/interview-cram - You already ship SW →
/roadmap/swe-to-ai(much faster) - You already know classical ML →
/roadmap/ds-mle-to-llm - You don’t code yet →
/roadmap/zero-to-ai
Estimated arc
40+ hours of focused reading isn’t cheating; it just spreads out. Most readers finish over 4-8 months of on-and-off Sunday afternoons. If you finish in 40 hours flat, you either skipped the exercises or you already knew most of it.
THE PATH
8 stops, in order
Phase 1Fundamentals2 stops
- 1
- 2
Track
1 · Classical ML in one weekClassical ML in one week. Enough to have opinions without becoming a data scientist.
Phase 2Core skills3 stops
- 3
- 4
- 5
Track
4 · Retrieval & Long ContextRetrieval and long-context. The most common production LLM pattern.
Phase 3Ship it1 stop
- 6
Phase 4Land the role2 stops
- 7
- 8
NOT COVERED HERE
What this roadmap skips
- Speed. This is the long path. If you have < 3 months, use `/roadmap/swe-to-ai` or `/roadmap/interview-cram`.
- Deep math. Foundations covers what you actually use.
- Model training from scratch. This is engineering around models.