Build the thing.
Understand why.
A working developer’s path into AI and machine learning, from your first messy dataset to systems and decisions you can defend.
Go from “it runs”
to “here’s the evidence.”
You know how to ship software. The next step is to learn what changes when a system learns from data: how to choose a test, read a result, and know when to keep the simpler approach.
Follow a practical order. Build something at every stage. Keep your decisions, failures and uncertainty in plain sight.
How to use this field guide ↗One thread.
A little deeper each time.
Eight core chapters. An applied business track.
A clear next step when you need one.
Getting oriented
Choose your direction. Set up a reproducible workspace. Make your first small commitment.
8 hoursWeek 002FoundationsData foundations
Turn messy records into a reproducible analysis, with a clear account of what you know.
50 hoursWeeks 1–503FoundationsCore machine learning
Build a baseline, protect the holdout, and connect a prediction to the decision it informs.
80 hoursWeeks 6–1304FoundationsDeep learning
Follow the gradient. Train a small network. Learn what to check when it fails.
90 hoursWeeks 14–2205EngineeringLanguage models & AI engineering
Use real examples and independent evaluations to build a useful language-model application.
100 hoursWeeks 23–3206EngineeringRunning it in production
Deploy, observe, recover and measure cost. Treat the whole release as a system.
55 hoursWeeks 33–3707Your practiceChoosing a specialisation
Commit to one role and build a capstone that shows your judgement.
70 hoursWeeks 38–4308Your practiceFinding the work
Show your results, prepare for interviews, and learn from a careful job-search log.
3–4 hours / weekAlongside Chapter 609Applied fieldworkBusiness machine learning
Connect offers, prices and retention to constraints, experiments and measured outcomes.
40 hours / caseApplied extensionPractice hours are planning estimates for people who already write software. The job search runs alongside the production chapters. See the full assumptions ↗
Pull on a thread.
See what changes.
Move a decision clock. Break a gradient. Change a price. Eight small experiments show where the important boundaries are.
Enter the laboratory ↗Or run the reference project locally →“It builds the foundation in the right order, and it teaches you to think in systems.”
Santosh Shinde — developer, author, fellow learner
Read the author’s note ↗