Part IX: Capstone Projects

"In theory, there is no difference between theory and practice. In practice, there is." — Yogi Berra (attributed)

The capstone projects in Part IX are designed to integrate concepts from across the entire book into substantial, production-grade applications. Each project requires you to combine knowledge from multiple parts — mathematical foundations, deep learning, transformers, and systems engineering — to build something that works end to end.

These are not toy exercises. Each project mirrors the kind of work that professional AI engineers do daily, and completing them will give you portfolio-worthy artifacts that demonstrate your capabilities.

Projects

Capstone Project 1: Build a Production RAG System with Guardrails

Chapters Applied: 20, 23, 31, 32, 33, 34

Build a complete retrieval-augmented generation system from scratch, including document processing, embedding, retrieval, generation, and guardrails for safety and quality. Deploy it as a REST API with monitoring and evaluation.

Capstone Project 2: Fine-Tune and Deploy a Domain-Specific LLM

Chapters Applied: 12, 19, 24, 25, 33, 34

Take a pre-trained language model, fine-tune it for a specific domain using LoRA, evaluate it rigorously, optimize it for inference, and deploy it with a serving framework. Include alignment evaluation and safety testing.

Capstone Project 3: End-to-End Multimodal AI Application

Chapters Applied: 26, 27, 28, 31, 32, 34

Build an application that processes both text and images — for example, a visual question-answering system or an image-grounded chatbot. Integrate retrieval, agent capabilities, and deploy with proper MLOps practices.

How to Approach These Projects

  1. Read the full project specification before writing any code
  2. Plan your architecture on paper, identifying which chapters provide the relevant techniques
  3. Build incrementally, testing each component before integration
  4. Document your decisions — the reasoning behind design choices matters as much as the code
  5. Evaluate rigorously — every project includes evaluation criteria and benchmarks

Prerequisites

  • Completion of Parts I–VI (at minimum)
  • Familiarity with Part VII topics is helpful but not required
  • A GPU-capable computing environment (local or cloud)

Chapters in This Part