12 parts
Go from 'what's an LLM?' to shipping a real AI app — prompts, structured output, tool calling, embeddings, RAG, and agents — in Python, provider-agnostic.
What a large language model really is: next-token prediction, tokens, training vs inference, and what LLMs are good and bad at, with a live demo.
10 min read
Make your first LLM API call in Python using the OpenAI-compatible SDK. Set up a client, keep your key safe, and get a real model response.
7 min read
How LLM prompts really work: the messages array, system/user/assistant roles, and keeping multi-turn conversation state in Python.
8 min read
Practical prompt engineering for developers: be specific, use few-shot examples, fix the format, and dodge the failure modes that waste tokens.
Control LLM output in Python: temperature, top_p, max_tokens, stop sequences, and streaming responses token by token, with an interactive demo.
9 min read
Get structured JSON out of an LLM reliably: JSON mode, a Pydantic schema, validation, and retries, so your code can trust the model's output.
Let an LLM call your Python functions with tool calling: define tools, run the call loop, and feed results back so the model can finish the job.
Embeddings turn text into vectors so you can search by meaning. Learn cosine similarity and build semantic search in Python, with a live explorer.
Build retrieval-augmented generation in Python: chunk, embed, retrieve, and ground an LLM's answers in your own documents with a working mini-RAG.
Build a simple AI agent in Python: the think-act-observe loop, tool use, and memory, so the model can take steps, not just answer once.
Tokens, cost, latency, and safety for LLM apps: count tokens, estimate cost, retry with backoff, and defend against prompt injection.
Capstone project: build a 'chat with your notes' AI app in Python with RAG, retrieval, and citations end to end, then where to go next.
12 min read
