Companies: how to get into the labs that build frontier AI¶
One guide per organization: where core-AI engineers actually work there, what the hiring process tests, the papers to read, portfolio projects that would impress that team, and a plan to get there. Each guide links back into the labs and tracks in this repo, so preparation is concrete.
Info
Team names, hiring processes, offices and programs change often. Every guide dates its volatile facts, and each folder has a refresh-brief.md to work through each quarter (alone or with a research assistant) to refresh them. Verify on the company's own careers pages before you act.
| Organization | What the guide covers | Start here |
|---|---|---|
| Anthropic | Interpretability, alignment, RL and pretraining, inference, applied AI; the values interview | Guide · Reading list · Projects |
| OpenAI | Research, post-training and reasoning, safety systems, scaling and inference, applied engineering | Guide · Reading list · Projects |
| Google (DeepMind, Research, Cloud AI) | Gemini, research engineering, JAX and TPUs, hiring committee and team matching | Guide · Reading list · Projects |
| Meta | Superintelligence Labs, FAIR, Llama, PyTorch; the E-level interview loop | Guide · Reading list · Projects |
| NVIDIA | Deep-learning software, CUDA and kernels, TensorRT-LLM, research; performance interviews | Guide · Reading list · Projects |
| More labs | Microsoft, Apple, xAI, Mistral, DeepSeek, Qwen, Hugging Face, Cohere, Ai2, India-based labs, startups | Guide |
How to use these guides¶
- Pick two targets, not ten. Read both guides end to end, then choose the projects that overlap. A kernel project, for example, serves NVIDIA and any inference team; an interpretability project serves Anthropic and research-heavy teams everywhere.
- Map the guide to your labs. Each guide says which labs and curriculum modules its interviews lean on. Finish those first, and turn one into a public write-up.
- Build the signal they look for. The "signals" section of each guide is ranked. Work down it in order, starting with what you can ship in the next 30 days.
- Prepare the loop. Use the research-engineer track for coding, system design and the question bank, and career/stories.md for behavioral and values rounds.
- Refresh quarterly. Work through each folder's
refresh-brief.mdand update anything that changed.
Tip
Before applying anywhere, have three things you can link: a lab you implemented and measured, a write-up with error bars, and one contribution to an open-source project the target team uses. That combination separates you from most applicants at every organization listed here.