From understanding to building.

The library draws its subject structure from the 32-week Advanced AI for Legal Practice syllabus. Use these lessons as a companion and a route into its practical builds.

What this library covers

The syllabus is a full-year, build-oriented programme. These shorter lessons explain the concepts and provide bounded exercises; they do not reproduce every multi-hour lab or constitute the completed course. In particular, training a neural network, production multi-GPU serving, and a complete litigation capstone require additional implementation and study.

Week	Syllabus subject	Start with this lesson
1	Computing fundamentals	Understand the hardware a local model needs
2	AI workstations and memory	Understand the hardware a local model needs
3	Python fundamentals	Your first useful Python script: a document inventory
4	Text and structured data	Your first useful Python script: a document inventory
5	Databases and SQL	Use a database to preserve facts and sources
6	Web and APIs	APIs and JSON: connect a model to a small program
7	Machine learning and neural networks	How models learn from examples
8	Tokens and attention	Tokens, attention, and why wording matters
9	Training, inference, and model evaluation	Why the same question can produce different answers
10	Local inference and quantization	Run a local model and verify the actual setup
11	Serving and multi-GPU concepts	Understand the hardware a local model needs
12	Prompting and structured assignments	Give AI a brief it can actually follow
13	Building a model-backed application	APIs and JSON: connect a model to a small program
14	Embeddings and retrieval	Keyword search, semantic search, and reranking
15	RAG fundamentals	How document-grounded AI answers work
16	Legal document ingestion	Prepare documents before asking AI to analyze them
17	Structured fact extraction	Build a chronology that survives checking
18	Tools and permissions	Build your first agent around a document inventory
19	Workflow and agent architecture	Prompts, skills, workflows, and agents
20	Agentic legal engineering	When do multiple agents help?
21	Evaluation	Test an AI workflow on work you can assess
22	Verification and adversarial testing	Recognize prompt injection in documents and tools
23	Fine-tuning and alternatives	Prompting, retrieval, or fine-tuning?
24	Persistent knowledge and context	Build a knowledge library you can actually reuse
25	Document review and discovery	Design a review table before generating it
26	Transcripts and examination preparation	Turn a transcript into useful examination preparation
27	Research and authorities	Research with AI without outsourcing the authority check
28	Evidence-led drafting	Build a drafting workflow around the record
29	Security and local/cloud boundaries	Map where information goes before using a tool
30	Deployment and operations	Turn a prototype into a service you can maintain
31	Litigation workspace integration	Make recurring work resumable and reviewable
32	Stress testing and rollout	Test an AI workflow on work you can assess

A useful capstone

Build a small fictional matter workspace that preserves original documents, retrieves relevant passages, stores reviewed facts, creates a source-linked chronology, supports one bounded workflow, and exports a verified work product. Add a known-answer test set, an access model, and a recovery procedure. Demonstrate the result to someone who did not build it.

Start with a small fictional record, then expand one dimension at a time. A prototype is complete for its stated scope when the actual outputs, checks, and recovery process work—not merely when a chat reports success.

Follow the technical path Find books and deeper courses