Intermediate / 6–8 weeks suggested pace

AI Applications

Understand the model. Build around it. Check the result.

Your path from learning to making

Explore tokens, sampling, temperature, and the difference between a plausible answer and a supported one.

Checkpoint: Explain why two runs may disagree.

Use public or your own non-sensitive documents. Split them into passages with stable IDs.

Checkpoint: Each passage can be traced to its source.

Implement a simple keyword baseline before experimenting with embeddings.

Checkpoint: Your test questions retrieve relevant passages.

If using a model API, pass retrieved context and require source references. Keep credentials out of your repository.

Checkpoint: An unsupported question receives an explicit limitation.

Check source relevance, unsupported claims, latency, and running costs.

Checkpoint: Publish a demo explanation and a table of observed results.

Browse existing courses and articles. Check each resource for prerequisites and its creator.