Skip to main content

Topic roadmap

Language models

18 lessons, about 50.5 hours in all, from the basics to advanced. Start with the first, or jump to what you need.

Lessons, easiest first

  1. Language models · Intermediate 1 · about 3 hCall an LLM API from codeSend a prompt to a language model from a Python script and use the reply in your program.
  2. Language models · Intermediate 1 · about 2 hErrors, retries and costHandle rate limits and API errors with retries, and put a cap on what your app can spend.
  3. Language models · Intermediate 1 · about 2 hGet structured JSON outputMake a model return JSON that matches a schema, and validate it before your code uses it.
  4. Language models · Intermediate 1 · about 2 hHow a model picks its next word: temperature, top-p and top-kExplain how a language model chooses each next token, predict what temperature, top-p and top-k do, and set them sensibly for a task.
  5. Language models · Intermediate 1 · about 4 hSystem prompts and prompt designWrite system prompts and instructions that give clear, repeatable results, and improve them by testing.
  6. Language models · Intermediate 1 · about 2 hTokens and the context windowExplain what tokens are, estimate the cost of a request, and keep inputs inside the context window.
  7. Language models · Intermediate 2 · about 2.5 hAI in Indian languages: tokens, quality gaps and what to useExplain why language models do worse and cost more in most Indian languages, measure the gap for your own users, and choose between a general model, translation and Indian-language models.
  8. Language models · Intermediate 2 · about 2 hHow a chat model is made: pre-training, SFT and preference tuningExplain the three training stages behind a chat model, say what each one adds and costs, and tell a base model from an instruct model when you pick one.
  9. Language models · Intermediate 3 · about 3 hTool callingGive a model a tool, run the call it asks for, and send the result back so it can answer.
  10. Language models · Pro 1 · about 3 hContext engineeringDecide what goes into a model's context on each turn: instructions, tools, examples, retrieved data, memory and tool results, and how to keep long tasks within budget.
  11. Language models · Pro 1 · about 3 hFine-tuning in practice: tools, settings and reading the loss curveRun a fine-tune sensibly: pick a base model, method and tool, set learning rate, batch size and epochs, read the loss curves, and debug a run that is not learning.
  12. Language models · Pro 1 · about 3 hHow fine-tuning works: data, training and what can go wrongExplain what fine-tuning changes inside a model, prepare a clean training set, judge a fine-tuned model fairly against a good prompt, and name its main risks.
  13. Language models · Pro 1 · about 2.5 hReasoning models and test-time compute: when to let the model think longerExplain chain of thought, best-of-N and reasoning models, pick the cheapest way to get a hard answer right, and keep the extra cost and wait under control.
  14. Language models · Pro 1 · about 3 hSynthetic data and distillation: training with model-made dataDecide when model-made data is worth using, generate and filter it safely, and distil a large model's skill on one task into a small, cheap model.
  15. Language models · Pro 2 · about 2.5 hModel merging: combining fine-tuned models into oneExplain what model merging is and when it beats training one multi-task model, describe averaging, task vectors, SLERP, TIES and DARE and layer stacking, and test a merged model safely.
  16. Language models · Pro 3 · about 6 hDesign an AI systemDesign an AI product on paper: its parts, how data moves, where it can fail and how you would measure it.
  17. Language models · Pro 3 · about 3 hPrompting, RAG or fine-tuningChoose between better prompting, RAG and fine-tuning for a problem, and explain the choice.
  18. Language models · Pro 3 · about 2 hQuality, cost and speed trade-offsExplain what you give up and what you gain when you trade quality, cost and speed against each other.

On these role roadmaps

Interview questions on language models

Books on language models

All recommended books

Today's AI turns are used

Lessons, roadmaps, videos, practice cards and the code playground stay free with your account, with no limit.