Market intel: targets, teams, formats¶
Where core-AI engineers work, which guide in this repo covers each organization, and how to research a specific team before you apply. Company-level detail (teams, hiring processes, reading lists, projects) lives in the company guides; this page is the map and the method.
Warning
Volatile. Last reviewed September 2026. Teams, locations, interview formats and compensation change often. Re-verify quarterly on careers pages, engineering blogs, Levels.fyi and in conversations with employees. Each company folder has a refresh-brief.md for its quarterly refresh.
Target map¶
| segment | organizations | guide |
|---|---|---|
| Frontier labs | Anthropic | companies/anthropic |
| OpenAI | companies/openai | |
| Google (DeepMind, Research, Cloud AI) | companies/google | |
| Meta (Superintelligence Labs, FAIR, Llama, PyTorch) | companies/meta | |
| xAI, Mistral, DeepSeek, Qwen (Alibaba) | more-labs.md | |
| AI infrastructure and hardware | NVIDIA (deep-learning software, CUDA, TensorRT-LLM, research) | companies/nvidia |
| Big Tech with India presence | Microsoft, Apple, Amazon, Adobe, and the Google, Meta and NVIDIA India teams | more-labs.md and the guides above |
| Model and platform companies | Hugging Face, Cohere, Ai2 and similar | more-labs.md |
| India AI builders | India-based labs such as Sarvam AI and Krutrim, and IIT/IISc-linked research groups; natural homes for Indic-language work | more-labs.md |
| Startups (founder-track learning) | early-stage AI companies where you own a whole system | more-labs.md; founder track |
Most frontier-lab core roles are in San Francisco, London or Paris, with some remote; the bar is global. For roles abroad, read visa-and-relocation.md alongside the guide.
Tip
Pick two targets, not ten, as the companies index suggests. Choose projects that serve both: a kernel or inference project serves NVIDIA and any inference team; an interpretability project serves research-heavy teams everywhere.
Interview formats (durable patterns)¶
These patterns have held for several years; confirm the current loop for each company in its guide and with the recruiter.
- Frontier labs lean on practical coding, ML deep dives, system design, and a values or mission conversation, with less emphasis on puzzle-style algorithm questions.
- Big Tech usually keeps data-structures-and-algorithms rounds, plus ML design and behavioral rounds tied to published principles.
- Almost everyone probes past projects deeply: every résumé claim gets tested (résumé guide).
Preparation: interview loops, question bank, stories.
How to research a team¶
- Careers page: the exact role text, location, level, and the team's name. Save a copy; postings disappear.
- Recent output: the team's papers (author lists name the people), blog posts, open-source repos and talks from the last 12 months.
- People: find 2–3 engineers on the team; read what they have published. This is also your outreach list.
- Tools they use: frameworks and infrastructure visible in their code and job posts. Contribute to one of them if you can (public presence).
- Conversations: one 20-minute chat with a current or former employee is worth more than hours of reading.
Per-team note (keep in your tracker):
Company · team · date researched · sources
What the team shipped or published in the last 12 months (links)
The problem they seem to be working on next
People I could learn from or contact
Which of my labs / write-ups is most relevant, and the gap to close
Interview format as reported (source, date)
Location, remote, visa sponsorship (source, date)
Compensation data¶
Use live data: Levels.fyi filtered by company, level and city; AmbitionBox for India; Blind with caution. Record the date and source of any number you rely on in your tracker; figures in older notes are stale. How to compare offers: offers-and-negotiation.md.