Unblocked, an AI context engine for software development, replaced a custom actor-model ingestion system with Temporal to handle continuous, durable ingestion across heterogeneous data sources. The team evaluated Prefect (ruled out for being Python-only), rolling their own (already doing that), and Temporal — which won on polyglot SDK support, native Durable Execution, and depth of capability. A thin Kotlin shim over Temporal's Java SDK collapsed pages of orchestration code to a few lines per workflow. Key gains include deployment-free long-running jobs, elimination of deploy-time DB spikes, unified cron and user-triggered flows, and built-in observability via the Temporal UI. Today Unblocked runs ~1 million Temporal Workflows per day with 3.6 million live workflows, self-hosted on Kubernetes with Aurora Serverless persistence and KEDA autoscaling.
