Interview
summary
Netflix runs depersonalization infrastructure that collects behavior from 100 million members and funnels it into both open-source and proprietary ML pipelines. The team standardizes streaming data for real-time jobs that must complete on micro-batches with a 10-minute SLA, yet failures in external services cause backlogs that cannot be backfilled. Solutions include end-to-end observability and fast triage; core tooling uses Scala and Spark MLlib and functional transformations for vectorized workflows.
AI-written summary · about
