Lucid Motors Software Engineer interview: PyTorch attention and C++ coding

My process started with an initial coding round in PyTorch. I had to implement multi-head attention from scratch, which was more hands-on than I expected.

After that, I moved into a panel-style stage with two technical interviews and two C++ LeetCode-style medium coding rounds. The pace was brisk, and the difficulty felt fairly consistent across the coding and technical discussions. Overall, it felt like a real test of whether I could work in both Python-based ML code and C++ problem solving. I left feeling they were looking for depth rather than rehearsed answers.

Location: Ann Arbor, MI. Overall feedback: Positive. Offer status: No offer.