AI Careers

How to Network Into an AI Company

AI labs get thousands of cold applications a week and most never get read. The honest path in is a warm intro from someone who already shares a lab, a school, or a past job with you.

Find Your Warm Intros →

Free · No sign-up · See results in ~10 seconds

Example — what you’ll see

Someone who works at your target company

🎓 Same university as you · 💼 Shared past employer

🔥 Strongest

A recruiter at your target company

🎓 Same university as you

🎓 Alumni

✍️ Ready-to-send intro“Hi — we both studied at [your school]. I’d love to hear about your path to a company you are targeting before I apply…”

… plus everyone else in your network who can put in a good word.

See who can refer you in — pick your target company:

Short answer: To network into an AI company, find people already inside who share something concrete with you (same university, past employer, open-source project, or research area), then reach out leading with that shared thing rather than a job ask. Ask for a 15-minute chat or a quick question about their team, not a referral up front. Referrals from real connections move you past the resume pile, which is where most cold applications die.

Why cold applications to AI labs almost never work

OpenAI and Anthropic receive far more applications than any recruiter can read, and the volume spiked once every engineer decided they wanted to work on frontier models. A cold resume competes against tens of thousands of others for a few open reqs, and there is no signal in it that separates you from anyone else who can list PyTorch.

A warm intro changes the math. When someone inside vouches for you, your application gets opened by a human instead of a filter, and the internal referral bonus gives that employee a real reason to pass your name along. You are no longer a stranger in a queue, you are a specific person a colleague chose to flag.

Find the shared thing before you write a single word

The mistake most people make is messaging strangers at the lab and opening with what they want. The people who actually get responses open with what they share, because that is the reason a busy engineer will spend 30 seconds on you instead of ignoring the notification.

Look for concrete overlap you can prove, then let it carry the message. Real shared context is the difference between a reply and silence.

Same school or lab. A person who did the same PhD program, took the same advisor, or graduated from your university feels a genuine pull to help, and it gives you an obvious first line that is not about a job.
Shared past employer. Someone who worked at a company you also worked at, even years apart, treats you as a near-colleague, and you can name specific teams or tools you both touched.
Open-source or research overlap. If you have filed issues, sent PRs, or read their paper closely, you can reference the exact work, which proves you are real and paying attention, not spraying messages.
Community and event ties. A mutual from a reading group, a hackathon, or a conference gives a mutual acquaintance who can make the actual introduction for you.

Let FindWarmIntros do the finding and ranking

The hard part is not knowing that warm intros work, it is finding which specific person at OpenAI or Anthropic actually overlaps with you and is worth reaching out to first. Combing through LinkedIn one profile at a time to spot a shared school or former employer takes hours you could spend on the actual conversation.

FindWarmIntros is a free tool that does that finding-and-ranking step for you: point it at a target AI company and it surfaces the real people there who share a school or a past job with you, ranks them by how strong the connection is, and drafts an opener that leads with the shared thing. You still send the message and build the relationship, the tool just removes the search.

Make the ask small enough to say yes to

Once you have the person and the shared context, the ask decides everything. Do not open by requesting a referral, because that asks a near-stranger to spend their reputation before they know you. Ask for something that costs them under a minute: one question about what their team is actually working on, or a 15-minute call about how they got in.

Referrals come after that first low-cost yes, not before it. When you have shown you are thoughtful and easy to help, offering to refer you becomes the natural next step, and often the person suggests it before you have to.

Frequently Asked Questions

Should I apply online too, or only try for a referral?

Do both, but treat the online application as a placeholder and the referral as the real move. Submit the application so there is a record in the system, then get someone inside to flag your name, because the referral is what gets that record opened.

What if I do not know anyone at the AI company at all?

You likely have more overlap than you think through school, past employers, or open-source work, which is exactly what surfaces second-degree connections. Start with people who share that context even if you have never spoken, since a real overlap beats a warm-sounding but empty message to a stranger.

How do I message a researcher without sounding like every other job seeker?

Reference something specific and true: their paper, a repo they maintain, or the program you both went through, and open with that instead of your resume. Then ask one small question rather than for a job, so replying costs them almost nothing.

Do referrals still matter for research or non-engineering roles?

Yes, arguably more, because research and specialist teams are small and hire heavily on trust and word of mouth. A vouch from someone on the team carries even more weight when there are only a handful of people making the call.

Keep going

Find Your Warm Intros →