Code Reference: The Project Catalog
The prose in this book teaches the why and the shape of each idea. The code — complete, runnable, tested — lives in a companion catalog of 52 from-scratch systems, linked throughout the chapters with the “Build it →” pattern. This appendix is the index to that catalog.
Each project implements one or more of the concepts a chapter develops, at production scale rather than illustration scale. Where a chapter says “Build it →”, it points here; where a project needs background, its README points back into these chapters. The full bidirectional map of which chapter ↔︎ which project lives in CONCEPT_TO_PROJECT_MAP.md.
All projects are under 06-real-world-projects/ in the repository. Python projects use FastAPI + pytest; Rust projects use trait-based design with Result-based error handling and Criterion benchmarks; the one Go system uses gRPC + protobuf.
Foundation & Backend (1–10)
Job queues, caching, microservices, data platforms, and ML training orchestration — the systems most chapters in Parts I, II, and IV build on.
Distributed Systems (11–20)
Consensus, networking, compilers, query engines, and SIMD analytics — the systems-programming tier, most heavily exercised by Part II’s Rust and C++ field guides.
ML / AI Core (21–37)
Embeddings, attention, RAG, model serving, parameter servers, and ML compilers. Part IV draws on the ML-systems members of this tier; the LLM/RAG members are the province of the companion AI Engineering book.
- 21 · Custom Embedding Model
- 22 · Long-Context Attention
- 23 · LLM Agentic Runtime
- 24 · Synthetic Data Generator
- 25 · RAG Baseline
- 26 · Advanced RAG
- 27 · Micro-Model Orchestrated RAG
- 28 · AI Workflow Engine
- 29 · Model Routing Layer
- 30 · Parameter Server
- 31 · ML Compiler
- 32 · Distributed Tensor Algebra
- 33 · RL Physics Engine
- 34 · Distributed File System
- 35 · Differentiable Programming
- 36 · Distributed Streaming Analytics
- 37 · Dynamic Graph Runtime
Advanced ML (38–49)
Distributed training, GPU memory and scheduling, inference engines, GNNs, and quantization — the heavy machinery behind Part IV’s Deep Learning, Distributed Training, and GPU chapters.
- 38 · Dynamic Graph Execution
- 39 · GPU Memory Manager
- 40 · Distributed Autograd
- 41 · Vector-Quantized LLM
- 42 · GNN Runtime
- 43 · Vector Index
- 44 · Autoregressive Inference
- 45 · Neural Compression
- 46 · Multi-Tenant GPU Scheduler
- 47 · On-Device LLM
- 48 · Multi-GPU Kernel Scheduler
- 49 · AI Benchmark Suite
Data Infrastructure (50–52)
The feature platform, message queue, and time-series database that anchor Part II’s Data Engineering chapters.
The catalog is broader than this book: several projects (notably the RAG, attention, and on-device-LLM systems, 21–27 and 41–47) implement concepts taught in the companion AI Engineering book rather than here. They are listed for completeness — this book links the ones whose concepts it teaches.