Resources
Everything the AI Engineering in 5 weeks track links to.
- [1hr Talk] Intro to Large Language Models
A one-hour, general-audience tour of what an LLM is, how it is trained and fine-tuned, where it is heading and where it can be attacked.
- Build Real AI Projects: From First Python Script to Autonomous Agents (Advanced AI playlist)
A code-along series that starts from a local setup and builds an assistant step by step, adding a conversation loop, memory and streaming. Good for typing along rather than watching.
- Building effective agents
Separates fixed workflows from agents, catalogues the common workflow patterns, and argues for the simplest design that works before reaching for autonomy.
- Deep Dive into LLMs like ChatGPT
A long, detailed walk through the full stack of a chat model, from pretraining data and tokenization to sampling, post-training and why models hallucinate. Watch it in parts.
- FastAPI documentation
Docs for the Python web framework suggested for the project backend, including streaming responses and request validation.
- Ollama
Runs open models locally behind a simple HTTP API, so you can build and test without paying per token.
- Prompt engineering interactive tutorial
Notebook-based exercises on prompting, from basic structure to formatting output and avoiding hallucinations. Do the chapters on output formatting for this track.
- Pydantic documentation
Docs for the Python data validation library used to define and check structured output schemas.
- Streaming messages
How a model API streams tokens as server-sent events, which is the pattern your chat endpoint will pass through to the browser.
- Your Roadmap to Master Advanced AI (playlist intro)
The short opening video of the Advanced AI playlist, explaining what the series builds and in what order.
- Contextual retrieval
Shows why chunks lose meaning when cut out of a document, and how adding short context to each chunk, plus BM25 and reranking, reduces failed retrievals.
- Effective context engineering for AI agents
Treats the context window as a scarce budget and covers compaction, note-taking and sub-agents for keeping long-running agents focused.
- pgvector
The Postgres extension for storing embeddings and running nearest-neighbour search next to your relational data.
- PostgreSQL: Full Text Search
The manual chapter on Postgres keyword search and ranking, the lexical half of a hybrid search.
- Build an MCP server
The official quickstart that walks through writing a small server, registering tools and connecting it to a client.
- How we built our multi-agent research system
An engineering account of an orchestrator that fans out work to parallel sub-agents, with the costs, prompting lessons and evaluation approach that came with it.
- Introduction to LangGraph
The official LangGraph course, covering graph state, persistence with checkpointers, and human-in-the-loop interrupts, with notebooks for each module.
- LangGraph documentation
Reference docs for graphs, persistence and memory stores; keep it open while rebuilding the agent.
- MemGPT: Towards LLMs as Operating Systems
Proposes managing an LLM's limited context like an operating system manages memory, paging information between fast and slow tiers. Skim for the memory model.
- Multi AI Agent Systems with crewAI
A short course on splitting work across role-based agents with crewAI. Optional contrast to the LangGraph approach.
- What is the Model Context Protocol (MCP)?
The official introduction to MCP, its hosts, clients and servers, and what servers can expose to a model.
- 12-Factor Agents
Principles for production LLM software, such as owning your prompts and control flow and keeping agents small and stateless where possible.
- Agent Harness Engineering
Argues that most of a coding agent's quality comes from the harness around the model, and that every mistake should tighten the harness so it cannot recur.
- AI Evals: Everything You Need to Know
Answers to common practical questions about evals, from error analysis and labelling to choosing metrics and when to automate.
- Langfuse documentation
Docs for an open-source LLM observability tool: traces across agent steps, token costs and prompt versions.
- OWASP Top 10 for LLM Applications
The community list of the most important security risks for LLM applications, each with examples and mitigations.
- Prompt injection (series)
A running series of posts on prompt injection attacks and defences. Read the earliest posts and a few recent ones.
- Ragas documentation
Docs for an open-source toolkit that scores RAG pipelines on things like faithfulness and context relevance.
- The lethal trifecta for AI agents
Explains why an agent with private data, untrusted content and a way to send data out can be tricked into leaking, and why removing one of the three is the dependable fix.
- Using LLM-as-a-Judge For Evaluation: A Complete Guide
How to build a model-based grader you can trust, starting from expert-labelled examples and measuring the judge's agreement with them.
- Your AI Product Needs Evals
Makes the case that evaluation is the core loop of AI product work, and lays out levels of tests from unit assertions to human and model review.
- The Illustrated Transformer
A diagram-heavy explanation of the transformer, with self-attention, multiple heads and positional encoding built up one step at a time.
- AI Engineering (book)
A book on building applications with foundation models. Its chapter on RAG and agents is a thorough companion to Week 2; the author's page links to where it is sold.