AI Engineer Path

If you fall behind

You will fall behind. Sundays are zero-hour by design to absorb it. If you are more than two weeks down, cut in this order, and never compress weeks 26–37 to catch up: that is the part that gets you hired.

  1. 1st

    Weeks 17 and 22: CNN depth, unsupervised DL

    Least downstream impact on the AI engineer track.

    Week 17: CNNs and transfer learningWeek 22: Unsupervised DL, deployment and review

  2. 2nd

    Week 10: unsupervised ML algorithms

    Skim instead of building.

    Week 10: ML algorithms (2/2)

  3. 3rd

    Weeks 23–25: Karpathy

    Fall back to 'Let's build GPT' + the tokenizer video, about 4 hours.

    Week 23: micrograd and makemoreWeek 24: Let's build GPTWeek 25: Tokenizer, nanoGPT and multimodal models

  4. Never

    Week 19 (attention), 31–32 (RAG), 34–35 (agents), 36 (evals)

    Everything depends on these.

    Week 19Week 31Week 32Week 34Week 35Week 36

Sundays are buffer, by design

The default pace leaves every Sunday at zero hours. They exist to absorb slippage and prevent burnout, so use them to finish the week rather than to start the next one.

Week 22 is deliberately light, with a buffer day. If you have slipped by then, absorb it there.

If you start the maths track early, weeks 4–5 become a review and give you most of two weeks of slack later.