Lessons
1How LLMs workNext-token prediction, tokens, attention, the training stages and the sampling knobs that shape every answer.8 min
2Tokens, context windows & costCount tokens, fit prompts into the context window and estimate the monthly bill before you ship.7 min
3Prompt engineeringStructure prompts with role, task, context, constraints and examples, then iterate against a small test set.8 min
4Structured outputs & tool callingGet reliable JSON from an LLM and let it call your Python functions safely via tool/function calling.7 min
5Embeddings & RAGTurn documents into vectors, retrieve the right chunks and ground an LLM in your own data with citations.8 min
6Building an LLM app with PythonWrap any provider behind one function, then add config, timeouts, retries, cost logging, caching and tests.8 min
7Evaluating LLM apps: hallucinations, RAG or fine-tuningMeasure quality with test sets, reduce hallucinations and choose between prompting, RAG and fine-tuning.7 min
8LLM security & building agentsDefend against prompt injection and data leakage, then build a safe tool-using agent loop.7 min