Tasks
All the tasks of the course in one place, by module. Each one leads to the lesson that explains what you need.
Tasks from the lessons
Section titled "Tasks from the lessons"How a model learns
- What machine learning is
- Linear regression
- Gradient descent
- Classification: fraud or not
- Evaluation: accuracy lies
- Overfitting: the model that memorises
A neural network from the inside
How a language model works
- Text in pieces: tokens
- Predicting the next token
- Embeddings: meaning as vectors
- The context window
- The limits of the model
Apps with LLMs
- Calling a model through an API
- The prompt is code
- Structured output: JSON by a schema
- Tool use: the model asks, the program acts
- Evaluation: a test set for the assistant
- Cost and latency: budget, cache, a smaller model
Semantic search and RAG
- Semantic search: the k nearest
- Chunking documents
- RAG: answers from the documents
- Evaluating RAG: does it find it, does it stand on it
Agents
- The agent loop: plan, action, observation
- Prompt injection: the email that gives orders
- Guardrails: checks before and after the model