Uber Data Scientist interview experience

A short while later, I had a 45-minute live screen that leaned heavily on advanced SQL window functions, plus some metric diagnostic questions where I had to reason about what the numbers would mean.

Then I moved into a virtual onsite loop that ran five rounds: product sense, stats and experimentation, applied modeling, data processing, and a behavioral bar-raiser. The stats/experimentation round felt like the real filter. I couldn’t get away with proposing a generic A/B test; they pressed on network effects and driver cannibalization, and they expected me to be comfortable with switchback