Flink operations and tuning heavily rely on expert experience, with slow anomaly resolution, high remediation costs, and continuous developer queries for SQL optimization and troubleshooting. To address this, we built Flink Ops, an intelligent operations platform integrating self-healing, diagnosis, and Q&A capabilities, and developed Flink Expert Agent powered by LLMs — forming a closed-loop system of “observe-decide-act-interact”. This session covers: Platform architecture: Unified ingestion of metrics, logs, lineage, and knowledge bases for real-time health profiling Self-healing engine: Dual-mode automation — rule-based remediation plus Agent-driven intelligent recovery Intelligent diagnosis: Automated anomaly detection, root-cause reasoning, and remediation recommendation Expert Q&A: Natural language interface for SQL optimization, configuration tuning, and troubleshooting Production results: MTTR reduced to X minutes, with Y% developer self-service rate Who should attend: Data platform engineers, SREs, and AI practitioners interested in intelligent DataOps and LLM-powered operational systems.

Speakers:

Chaoming Zhang: Senior Engineer, Ant Real-Time Computing Team

6 years in big data. Senior Engineer, Ant Flink Engine Team. Building Flink Ops and Expert Agent for intelligent real-time computing operations.