Three interviews spanned the 2nd and 3rd week of May. Ultimately, I did not receive an offer, but the experience was quite positive and an excellent learning experience, especially for a non-CS person looking to break into the world of autonomous driving. Ms. Song, a Senior Recruiter at Waymo, was my point of contact and guide throughout the entire process.

Interview 1: An informal project with a senior software engineer. They provided details on the newly formed Multi-System group, followed by information regarding the role, responsibilities, and learning opportunities of this internship position. The interview concluded with a discussion of my Ph.D. project. Due to NDA, I am unable to share any further details.

Interview 2: C++ coding and algorithms. A code-along programming exercise involving graphs and the use of a suitable graph traversal technique. The problem presented was very similar to LeetCode problems but required some thinking.

Interview 3: A general domain knowledge test, which unexpectedly included two additional C++ coding exercises. The interviewer was a Ph.D. in Geometric Computer Vision from a very prestigious university. I likely faltered during this part of the interview, mostly due to my not expecting the implementation of vector products and a form of nearest neighbor problem. Due to NDA, I cannot share any further details.

None of the three C++ coding exercises assumed that the code needed to compile and run. The test, I believe, focused more on my command of the language, algorithms, and ability to assess verbal/written problems and think quickly in a high-pressure interview scenario. Additionally, both interviewers asked to write interfaces (simple functions) and expected the use of simpler data structures. I made a mistake by trying to use too much Modern C++ (smart pointers, vectors, classes).

A code-along programming exercise involving graphs and the use of a suitable graph traversal technique.

A problem related to calculating surface normals given a set of points.

A problem related to point clustering around a query point. There was a twist, which due to NDA, I am unable to share.