AI engineer
Building real products on top of language models: retrieval, agents, evals and serving at scale.
What does a ai engineer do?
An AI engineer ships software that uses models, rather than training them. The job is mostly engineering: data in, structured answers out, and knowing when the answer is wrong.
Every stage below has chapters in the AI engineering path, and the last one is where it meets system design: serving, caching and rate limits.
- Stage 01Foundations
Python and TypeScript are both first class for AI work now. Pick the one you already write and get fluent in it.
3 of 4 writtenPython or TypeScriptAsync and await3 min - Stage 02How models work
Enough of the inside to predict the outside.
4 of 4 written - Stage 03Building with models
Turning a model into a feature.
5 of 5 written - Stage 04Retrieval
Giving a model the facts it does not have.
6 of 6 writtenCheckpointDesign a support assistant - Stage 05Advanced retrieval
When one search is not enough, and when documents are not just text.
4 of 4 written - Stage 06Agents
Models that take steps, and the rails that keep them safe.
8 of 8 writtenCheckpointDesign a coding agent - Stage 07Tools and protocols
How agents reach your systems and each other, safely.
6 of 6 written - Stage 08Serving at scaleCheckpointDesign an LLM inference platform
The same system design problems, with a GPU in the middle.
5 of 5 written