AI Lab Referrals

Warm Intros to Researchers at Anthropic and OpenAI

Research and research-engineer roles at top labs get thousands of cold applications and almost no warm ones. If you share a lab, a school, or a past team with someone inside, that is the door most applicants never knock on.

Find your warm intro →

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: Find a researcher at the lab who shares your school, past employer, or a co-authored paper. Message them about the specific thing you share and one concrete detail of their work, not your resume. Ask for a 15-minute chat, not a referral, and let the referral come up naturally once they see you are legit. A referral from a researcher routes your application to the hiring team instead of the cold pile.

Why a warm path matters more at research labs

Labs like Anthropic and OpenAI hire researchers and research engineers on judgment and taste, which is hard to read off a resume. A note from an internal researcher who has talked to you is treated as a signal that a keyword-matched application never carries.

Cold applications to these roles compete against people with published papers and lab experience. A warm intro does not fake that record, but it gets a human to actually read your work and your GitHub instead of a filter rejecting you in four seconds.

How to reach the researcher without getting ignored

The mistake is opening with the ask. A researcher gets pinged constantly, so make responding cost them 30 seconds and make the shared thing impossible to miss.

Lead with the overlap, not the request. Open the first line with the school lab, the past employer, or the paper you both touched. That is why they will read past the first sentence.
Name one specific thing in their work. Reference an actual result, method, or blog post of theirs and one honest question it raised for you. Generic praise reads as a template.
Ask for a 15-minute chat, not a referral. A chat is a low-cost yes. Referrals come after they believe you can do the work, so earn that first in the conversation.
Attach proof they can skim. Link one repo, paper, or writeup that shows you can do the kind of work they do, so the pitch is your output, not your adjectives.

Finding who actually shares a connection with you

The hard part is knowing which researcher at the lab you have a real thread to, because org charts and LinkedIn searches bury it. This is the part FindWarmIntros does for you: it looks at a target lab, surfaces the people there who share your school, a past employer, or an overlap with your history, ranks them by how warm the path really is, and drafts an intro you can edit. You still send a human message, but you skip the hours of guessing who to write.

Once you have the name and the shared thread, the connection is yours to make real. The tool finds the door; the conversation is still on you.

Frequently Asked Questions

What if I have no papers and I am applying for research engineer roles?

Research engineer roles weigh strong engineering and systems work heavily. Lead with a repo or a project that shows you can build and debug real training or inference code, and be honest that you are strong on the engineering side. A shared ex-coworker who saw you ship can vouch for exactly that.

Is it weird to message a senior researcher I never met?

No, if the shared thing is real and specific. Alumni and former colleagues expect these notes. What reads as weird is a vague message that could have been sent to anyone, so anchor it to the one concrete overlap you actually have.

Should I ask for the referral in the first message?

Usually not. Ask for a short chat first. If the conversation goes well and the fit is clear, either they offer to refer you or you ask at the end. A referral given after a real talk carries far more weight than one extracted cold.

How many researchers should I reach out to at one lab?

Two or three with a genuine shared thread beats twenty cold ones. Prioritize the closest connection first, personalize each note, and follow up once after about a week if you hear nothing. Depth of the shared signal matters more than volume here.

Keep going

Find your warm intro →