I applied through other source. The process took 2 months. I interviewed at Meta (Menlo Park, CA) in Jan 2025
Interview
There were two rounds of interview, first was programing and then research. Programing questions were quite easy, but since I am not used to be looked at while I am coding, it was still difficult to do it quickly. Res
Interview questions [1]
Question 1
I was asked to implement one deep learning question in (pseudo)PyTorch and then one data structure. For the research, I first talked about my research and then I was asked one research question regarding Llama.
I applied online. The process took 2 months. I interviewed at Meta
Interview
There was 1 research and 1 coding interview. The research interview consisted of 2 parts: in the first part, they asked questions about my publications, while the second part was a brainstorming session about a research problem.
I had two interviews: one for research fit, where I presented my own research and the interviewer asked a lot of open-ended questions and the second was 45 minutes of ML coding (no pytorch), which was not very hard.
there is one coding round, just two lc easy questions, and 1 research round, asked about my research papers, gave hypothetical situation and asked for specific model architecture design details. Overall was pretty normal and not hard.
Interview questions [1]
Question 1
1. Dot Product of Two Sparse Vectors
2. Randomly sampling K elements from a list without replacement
