J'ai postulé en ligne. J'ai passé un entretien chez Anthropic
Entretien
Screening - recruiters prep material was on qkv attention, which I studied and learned how to do in pytorch from scratch. I could probably do it straight from memory at this point. Then the interviewer only asked about attention free transformers. What was the point of giving me prep instructions that would never be used?
I don't feel I was able to display any kind of ML knowledge or any other kind of skillset during this interview. All this showed me was they're looking to over index on someone who can quickly figure out matrix manipulations and not much else.
My interviewer also had no ML background from what I could see on his LinkedIn.
J'ai postulé en ligne. J'ai passé un entretien chez Anthropic en mai 2025
Entretien
1. Recruiter call
2. Pre-screening on CodeSignal
3. Video code-screening
4. Onsite (4 interviews: 1 code, 1 SD, 2 behavioral)
5. References
The references stage is not a formality. They seem to take it very seriously. think very deeply about who you put down if you make it this far.
J'ai postulé via un recruteur. Le processus a pris 2 semaines. J'ai passé un entretien chez Anthropic (San Francisco, CA) en avr. 2024
Entretien
Got asked to do pre-screening coding round on Codescreen, and was offered a retake (perhaps because I was borderline) and then was rejected even after scoring nearly perfect score on retake. The recruiters are generally helpful during the process and are quick to respond to the candidate's queries.
Questions d'entretien [1]
Question 1
Similar to a software development assignment with a bit of leetcode-style problems baked in.
