Applied CS — Projects
The companion code for the Applied Computer Science book — 52 from-scratch systems in Rust, Python & Go spanning distributed systems, ML/AI infrastructure, databases, and language/OS internals. Each is honest about what's real vs. simulated.
What This Is
The companion code for my book Applied Computer Science — 52 from-scratch systems that implement the concepts the book teaches. Where a chapter says "Build it →", it points here. Distributed systems, ML/AI infrastructure, databases, and language/OS internals, written in Rust, Python, and Go.
Browse the repo → github.com/jchu0/applied-cs-projects · Read the book → jameshu.io/books/applied-cs
What's Inside
52 self-contained projects, each built to real engineering standards — trait/interface-driven design, typed errors, and a substantial test suite — and each honest about its scope. Some are complete, hardened services; others are deliberately CPU-only simulations or teaching implementations (the GPU, OS-kernel, and async-runtime work in particular). Every project's README and docs/BLUEPRINT.md state what is real versus simulated, so nothing overclaims.
- •Foundation & backend: distributed job queue, microservice platform, redis-lite cache, ML training orchestrator, SaaS platform, async runtime, data lakehouse, streaming, data observability, semantic layer.
- •Distributed systems: Raft key-value store, Kafka-lite log, service mesh, TCP/HTTP network stack, minimal OS kernel, CRDT engine, columnar query engine, a Python-subset compiler, GPU GEMM, and a SIMD analytics engine.
- •ML / AI core & advanced ML: custom embeddings, long-context attention, an agentic runtime, RAG baselines, a parameter server, an ML compiler, distributed autograd, a vector-quantized LLM, an autoregressive inference engine, a multi-tenant GPU scheduler, and more.
- •Data infrastructure: feature-engineering platform, message queue, time-series database.
How It's Built
- •Python (34 projects) — FastAPI +
pytest, packaged withpip install -e .. - •Rust (17 projects) — trait-based design,
Resulterror handling, Criterion benchmarks, built withcargo build/cargo test. - •Go (1 project) — gRPC + protobuf.
Every project ships its own README.md, a docs/BLUEPRINT.md design document, and build/run instructions. Served APIs share an opt-in hardening baseline — API-key auth, in-process rate limiting, and request timeouts.
Who It's For
Engineers who learn by building the real thing rather than reading about it — used alongside the Applied Computer Science book, or standalone as a reference set of production-shaped implementations across six problem domains.