Maintain large kafka clusters, futz with mesos, and futz with kubernetes. Same as all of us really. The difference is that a lot of startups will talk about the

Maintain large kafka clusters, futz with mesos, and futz with kubernetes. Same as all of us really. The difference is that a lot of startups will talk about their "hard big data problems" and might have 500T. A firm like Bloomberg's "big data problems" might be 400+ Petabytes. They also generally speaking totally shun the cloud as at that scale it tends to be cheaper to build on-premise.