There were four round interviews, consisting of three tech interviews and one behavioral interview.

The first of the three tech interviews were held in three consecutive weeks. The behavioral interview was scheduled two weeks after the last tech interview.

All interviews were conducted via phone or WeChat.

The scope of questions was very broad. I recall questions on matrix completion, MCMC, simple Bayesian reasoning (e.g., coin flips with a prior), SVM, variational inference, gradient boosting trees, and a bit of deep learning knowledge.

Besides these, I was asked to describe one of my projects, followed by two specific questions: What is the drawback of my model, and how can I improve it?