The interviews mostly focused on past systems experience. They had me describe the architecture of a couple systems I'd previously built, and then dug into the details, tradeoffs, and lessons learned for a long time. Additionally there were a couple of short whiteboarding questions, one that was all about probability and another that focused on algorithm design. It seemed like the most important part of these was actually thinking through them, asking good questions along the way, and collaborating with the interviewer.
