AI made easy with Flink + AIFlow
summary
AIFlow implements an event-driven workflow layer that turns logical AI nodes into executable jobs, chaining preprocessing training validation and serving into a single Flink job when possible and otherwise emitting engine-specific jobs for Python Spark or TensorFlow. It uses event-driven scheduling, a model center + notification service and a metadata store so long‑running streaming trainers can publish model versions that trigger validation and instant model reloads on Kubernetes or other platforms. The demo shows it working end to end with Flink.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:47Agenda
- 1:49Data
- 10:20AIFlow
- 14:05Model Center
- 19:12Execution Blocks
- 21:53Example
- 26:53Flink AIFlow
- 30:56Recap
- 31:45Demo
- 36:20Summary
related talks
37mBuild a Flink AI Ecosystem
36mInfusing Trusted AI using Machine Learning Payload Logging on Kubernetes
38mAI in Kubernetes: How to Get Started?
40mFrom concept to production; the fastest way to develop & deploy Apache Flink jobs
42mAI – Addressing the Challenges of Today’s Data Scientists
53mIntroducing AI to the Flowable Process Engines
54mArtificial Intelligence seen from the software development lifecycle perspective
27mAI-Powered ETL Pipeline Orchestration: Multi-Agent Systems in the Era of Generative AI