About

Learn Backend teaches production AI engineering to experienced backend engineers. Not "what is an API": you already ship those. The premise is that LLM systems are backend systems with new failure modes, and the fastest way to learn them is mapped onto what you already know: retrieval is search infrastructure, evals are your test suite, token budgets are resource limits, and a model provider is a flaky upstream dependency.

Everything here is written by a backend engineer working at a fintech, where LLM calls sit in real request paths with real latency budgets and real bills. The site is credited as a publication rather than a personal blog. The work should stand on whether the code runs and the numbers check out, not on whose name is attached.

What's here

The AI Engineering track: a free, structured curriculum from LLM fundamentals to RAG, agents, evals, and serving. Ships roughly one module a week, in order.
Guides: standalone decision guides and comparisons for the questions you hit mid-build: which vector store, how to stream, when to cache, which eval tool.
The blog: essays and production notes that don't fit the curriculum.

Editorial standards

Code is run before it ships. Every runnable sample is executed against the current provider APIs before publishing.
Guides carry a "last verified" date. When pricing, model names, or APIs change, the guide gets re-checked and the date moves, or the guide says so.
Comparisons are vendor-neutral.No affiliate links, no sponsored placements, no pay-to-win rankings. When we recommend a tool, that's the whole transaction.
Trade-offs over hype.Every recommendation includes when it's the wrong choice.

Contact

The best way to reach us is to subscribe below and reply to any issue. Replies land in a real inbox and get read.