Anthropic interview experience

Of all the places I've interviewed, Anthropic has by far been the weirdest. I applied for an RE position after one of their recruiters reached out to me on LinkedIn. I work on pretty related areas at Google and I believe this is why they decided to interview me. We skipped the codesignal round and jumped ahead to the ML round.

The ML round had an interesting problem on binary classification with LLMs. The recruiters share a two-line description with topics to help you prepare. Just be familiar with basic ML concepts (like logistic regression, classification, etc.). The ML round went well and got positive response within a few days. Next, they wanted to arrange a chat with a hiring manager for the multimodal team. The HM did not seem very experienced in this area at all. No prior work or publications in anything multimodal or LLMs related. They seemed to have joined Anthropic straight out of graduate school with prior work unrelated to ML.

Asked me bunch of generic questions such as "why Anthropic", etc. No technical questions to judge depth, etc. Rejected a few days later with a generic reason and no specific feedback. Pretty disappointing but secretly feel like I might have dodge a bullet here.

The recruiters were super nice and accommodating throughout. They deserve a raise! Overall, the impression that I got after speaking to at least 3 of the interviewers is that the company has grown quite aggressively in the last 12-18 months. A lot of this growth is through hiring people without a strong experience but those who can do research (e.g., they hire a lot of physics PhDs with limited ML experience). This is generally fine for ICs since good researchers can pick up LLM stuff pretty quickly but for managers overseeing fairly large groups I think this leads to a lack of strong leadership and vision.

This may just be me but I definitely felt weird after meeting both the interviewers. One of them told me that they don't personally believe in publishing because "AGI is only 2-3 years away". They also repeatedly bragged that they put in long hours at work which just seemed like an obvious red flag.

Most researchers working in this field at top labs are excited about the rapid progress happening lately and are naturally quite happy to work hard but to brag about it as it somehow directly relates to productivity is just odd to me. What's next? A d**k-measuring contest based on who wrote the most lines of codes this quarter?

Another weird vibe I picked up from the interviewers is a sense of entitlement. Like they have a moral chip on their shoulder for all the safety work that Anthropic does. It is great that Anthropic has a strong safety focus but the superiority that some of the researchers seemingly project to outsiders is quite off-putting and doesn't help their case as they're looking to hire and grow aggressively. But, hey, this may just be me. I could be misinterpreting the air here. I don't believe that I am because I interviewed at several other places where I did not get this kind of vibe.

However, I will say that this did not apply for the recruiters, who were very sweet and professional. One of the things that the recruiter mentioned is they find that people do well on the technical interviews but fail to pass the final "behavioral" round which is heavily safety-focused which they take very seriously. Read the latest Anthropic papers on safety and be prepared to show some depth and meaningful interest in this area. Also, I think it helps if you believe that "AGI is only X years away" where X <= 5. They won't make you bet on it but somehow this absurdly specific belief will convince them that you're one of them and they may let you in to their cult :)

#Anthropic, #genai