I went with a typical SQL based architecture. User->Job tables. Every minute scan for jobs that need to be done. Pawn off the work to workers. Update the job when done. The prompt they gave was how I would scale it up to higher workloads, like 10 jobs per minute. Which was kind of ludicrous because 10 jobs/min is trivial. I'm sure a raspberry pi could do 100x that. I clarified if they meant 10000 jobs per min, or some arbitrarily high amount, the interviewer said sure. So then I added some job queues and said to switch out the single db with a nosql database. The interviewer did not really dig into much because they said they were from the data side and wasn't in backend.