Netflix's Two-Tier Workflow Framework: RAW Hollow for State, Conductor for Execution

🎬 A new take from JSNation Amsterdam, June 2026 - your pipeline treats human review as an exception. Netflix made it a worker type: the same slot rules, ML models and AI agents plug into. What made that possible wasn't AI at all 👀 What's inside: 🧩 Two custom workflows before this one - 7 states in one, 9 in the other. Both scrapped. 🔁 A human correction isn't a note. It's structured output that flows straight into model retraining. 🔌 Write a sub-workflow and a config - dashboards, QC and human review come for free. 🎙️ Pratyusha Singaraju, Senior Software Engineer at Netflix, builds the backend behind content understanding — the systems powering search, collections and ratings for millions of viewers. Previously on Microsoft's Bing data team. ▶️ Full talk on our YouTube channel 🤍 #JSNation#GitNation #Netflix #NetflixEngineering #SoftwareArchitecture #SystemDesign #Backend #BackendEngineering #DistributedSystems #Orchestration #WorkflowOrchestration #ConductorOSS #NetflixConductor #RAWHollow #Hollow #StateManagement #EventDriven #EventDrivenArchitecture #Microservices #PlatformEngineering #DeveloperPlatform #InternalTools #HumanInTheLoop #HITL #MLOps #MachineLearning #MLInfrastructure #ModelRetraining #FeedbackLoop #AIAgents #AgenticAI #AIEngineering #ContentMetadata #Metadata #DataEngineering #ScalableSystems