拿到DeepMind、Meta等多家Offer后，她把机器学习求职面试的真相写了出来

TL;DR - A new DeepMind research scientist shares a candid guide to landing top machine-learning roles, based on successful interviews with DeepMind, Meta, Cohere, Isomorphic Labs, and a startup. It matters because it exposes both the technical expectations and psychological strain often hidden behind these hiring processes.

• Strong candidates may need 3+ first-author papers and relevant internship or industry experience just to secure interviews.
• Preparation emphasized roughly 150 medium-level coding problems and timed implementations of Transformers, attention variants, Flash Attention, backpropagation, and training loops.
• LLMs such as Claude or Gemini can simulate role-specific interviews; the author found their questions often overlapped with actual interviews.
• Coordinating interview timelines and competing offers improves negotiating leverage, while startup equity should be discounted for tax, liquidity, dilution, and expiration risks.