Waymo VO Rejected - sharing process for anyone prepping

Just finished the full loop at Waymo and got rejected. Sharing everything I can remember since there's honestly not much on the data fluency rounds online, and I wish I had found something before I bombed it.

Background: 5 YOE, currently at a mid-size infra company. Recruiter reached out on LinkedIn. Phone screen scheduled about a week later.

Phone Screen (45 min) Graph shortest path on a chess board. Fixed board size, find shortest path from A to B with some cells blocked. I used BFS which the interviewer seemed fine with. Follow-up was to handle blockage updates dynamically. Felt okay honestly, didn't think I'd pass but they moved me forward.

Virtual Onsite (4 rounds)

Coding (round 1): Given a custom graph where nodes are "waypoints" and edges have weights, find the path with minimum effort. Basically LC 1631 but wrapped in a different story. Had to define my own input format which threw me a bit at first. Wrote it in Python, got through the base case and one follow-up.

Coding (round 2): Sort a string by a custom priority order. Finished clean, felt good about this one.

System Design: How would you evaluate a self-driving model? Super open-ended. No diagrams, no rubric, the interviewer just started asking edge-case questions. Stuff like: what if simulation data doesn't reflect real-world distribution? How do you compare two model versions when the safety metrics conflict? I thought this went well but honestly it's hard to tell with Waymo SD rounds.

Data Fluency: This is where I got cooked. They asked me to design a statistical experiment to measure whether a new planning module reduces latency. I know A/B testing at a surface level but when they pushed on sample size, statistical power, how to handle non-normal distributions in latency data... I just didn't have the depth. Two questions, both in this category, and I was winging it both times.

HM round was standard BQ, nothing surprising.

Got the rejection email a week later. Automated. No feedback.

Data fluency prep is clearly where I dropped the ball. Gonna actually study experimental design and latency analysis properly before the next one.

YOE: 5