Prerequisites
Solid Python including async, decorators, and package management. ML fundamentals (loss functions, gradient descent, overfitting). Basic cloud/DevOps concepts (Docker, REST APIs). Chapters 0 through 2 can be skimmed as review.
Focus: Full stack from APIs through production deployment. Students leave able to architect, build, and operate LLM systems at scale. This pathway assumes students already have solid ML fundamentals, so it starts with a transformer review and quickly moves to the practitioner stack. The ordering (inference optimization before APIs, fine-tuning before RAG, agents before production) mirrors the typical system design sequence.
14-Week Syllabus
| Week | Topics | Lab / Assignment |
|---|---|---|
| 1 | Transformers and Decoding (review, Ch 04 through 05) | Implement a transformer block with KV-cache |
| 2 | Pre-training, Scaling, Model Landscape (Ch 06 through 07) | Analyze compute-optimal training configurations |
| 3 | Inference Optimization | Profile and optimize inference latency |
| 4 | APIs and Prompt Engineering (Ch 10 through 11, incl. DSPy) | Build a production prompt management system |
| 5 | Hybrid ML+LLM Architectures | Design an ML+LLM pipeline for a real use case |
| 6 | Fine-Tuning and PEFT (Ch 14 through 15) | LoRA fine-tune a 7B model on domain data |
| 7 | Embeddings, Vector DBs, RAG (Ch 19 through 20) | Build a production RAG pipeline with evaluation |
| 8 | Conversational AI | Build a task-oriented dialogue system |
| 9 | AI Agents and Tool Use (Ch 22 through 23) | Build an agent with MCP tool integration |
| 10 | Multi-Agent Systems and Agent Safety (Ch 24, 26) | Implement supervisor and debate patterns |
| 11 | Evaluation and Observability (Ch 29 through 30) | Build an LLM evaluation harness |
| 12 | Production Engineering and LLMOps | Deploy with CI/CD, monitoring, and alerting |
| 13 | Safety, Ethics, Strategy (Ch 32 through 33); further reading: Emerging Architectures (Ch 34), AI and Society (Ch 35) | Red-team an LLM application; write a risk assessment |
| 14 | Final project presentations | Production-grade LLM system (team project) |
- Appendix D: Environment Setup – set up your development environment before Week 1
- Appendix K: HuggingFace: Transformers, Datasets, and Hub – access pretrained models and datasets for labs
- Appendix S: Inference Serving – deploy and serve models at scale
- Appendix V: Tooling Ecosystem – survey the broader tooling landscape for production systems
What Comes Next
Return to the Course Syllabi overview to explore other courses and reading tracks, or proceed to FM.4: How to Use This Book for a quick orientation on conventions and callout types.