Start with the Python course embedded next in this roadmap. It begins with running a script and ends with a tested document-processing application. You can also open the same course as a standalone Python roadmap; the lesson content is shared while progress is tracked within the path you use.
Continue through software, data, and mathematical foundations before the AI application modules. Build the exercises and keep their outputs. The first project processes documents without a model. The later project adds retrieval, generation, evaluation, and deployment to a familiar problem.
Everything is self-paced. There is no weekly schedule or cohort deadline. Existing programmers can inspect the Python exit checklist and skip material they can already demonstrate, while beginners should follow the sequence. Marking a lesson complete records study progress, not a professional credential.
Exercise: create a learning repository with folders for notes, exercises, evaluation data, and project evidence. Write a short explanation of your current experience and the first concept you need to learn.
Check: you know where to start, what you will build next, and how you will demonstrate that a stage is complete.
Your project sequence
Treat each module as a cycle: read the concept lessons, run the guided example, attempt the independent exercise, and compare your evidence to the checkpoint. The short concept pages and longer workshops work together. Do not skip the workshop because its title begins with “Lab.”
| Stage | Artifact you keep | Evidence that you can move on |
|---|---|---|
| Python | Document-processing CLI | Invalid records, replay, updates, and persistence tested |
| Software and math | Reproducible application and experiment notes | Request flow, isolation, statistics, and data splits explained |
| AI applications | Validated classifier and retrieval baseline | Separate schema, label, and evidence-quality results |
| Production | Release scorecard and operational runbook | Access tests, outages, disable, and rollback rehearsed |
| Capstone | Independent knowledge assistant | A peer reproduces the build and challenges the evidence |
Use Python 3.10 or newer for the full course. Most local examples use only the standard library. SDK, model-provider, browser, container, and cloud exercises state additional setup where introduced. API credentials are optional for fixture-based learning but required for a real-provider evaluation; cloud deployment requires an environment you control. Never count a fixture run as a live-model or deployed-service result.
Placement check: an experienced developer can start by implementing the Python exit project from its requirements without copying the sample. If input validation, SQL transactions, or async cancellation is unfamiliar, take those modules before model integration. A beginner should follow the existing order.
Self-assessment: after a module, explain the example without running it, predict a failure, implement the challenge independently, and preserve the result. Marking the page complete records your progress; the saved artifact demonstrates the skill. The final connected build shows how the pieces fit before you complete the independent capstone.