First Principles Engineering
Writing about software engineering foundations, system design interviews, and AI systems — built from the ground up, not from buzzwords. Curated notes I'm willing to be wrong about in public.
Start with a roadmap
Roadmap
Foundations Roadmap
For newer engineers and gap-filling: networking, APIs, databases, caching, distributed systems, and production basics.
Start here
Roadmap
System Design Interviews Roadmap
For interview prep: turn requirements into APIs, storage, caches, queues, scale, reliability, and trade-off narratives.
Interview prep
Roadmap
AI Systems Roadmap
Understand the engineering around the model: data quality, retrieval, serving, evaluation, cost, latency, and observability.
Growing roadmap
Recommended Reads
AI Systems
RAG Architecture
The most-deployed LLM pattern in production is mostly a retrieval system with a model bolted on — not the other way around.
Databases
MVCC
Uber's 2016 migration from Postgres to MySQL forced the community to reckon with what "multi-version" actually costs in practice.
Distributed Systems
Consensus Algorithms
Paxos, Raft, and why Lamport spent eight years arguing about Greek allegory before "Paxos Made Simple" finally shipped.
Caching
Consistent Hashing
The textbook ring is a classroom curiosity. Virtual nodes are what makes it work in production — and most explanations skip them.
Distributed Systems
Logical Clocks
Lamport imported special relativity into distributed systems because physical timestamps quietly lie under clock skew.
Distributed Systems
Distributed Transactions
Two-phase commit, Jim Gray's disappearance, and the cruel irony of a community that couldn't coordinate finding its own founder.
Data
Change Data Capture
Why dual writes always drift. CDC reframes the database write-ahead log as the source of truth — not an implementation detail.
APIs
gRPC vs REST
REST works until JSON parse cost dominates your CPU bill and bolted-on WebSockets become load-bearing. Then it doesn't.
Networking
Load Balancers
Round robin is the easy part; the production lesson is that health checks, connection stickiness, and L4/L7 choices decide the outage.
Networking
HTTP/1.1, HTTP/2, and HTTP/3
The API looks the same until TCP head-of-line blocking, multiplexing, and QUIC explain where the latency really went.
About this project
Learning in public
First Principles Engineering
Curated notes from a larger private notebook
Start with a roadmap when you want structure. Use search when you already know the concept. The site intentionally keeps learning paths few and clear so the homepage does not become another notes folder.