Adding Generative AI to Real Time Streaming Pipelines
summary
Shows a working real‑time pipeline that ingests Slack messages and images, routes them through Apache NiFi 2.0 Python engine processors, pushes events to Kafka, enriches with Flink or Ray, vectorizes to a vector DB and runs vector DB approximate nearest neighbor searches. Demos include image captioning and YOLO/ResNet50 classification, Slack queries resolved to nearby bus routes (Trinity College Dublin, Boston), Spanish translation, and recorded token counts and latency; the pipeline works end‑to‑end and logs provenance for debugging.
AI-written summary · about
chapters (12)
- 0:00🚀 Introduction to Streaming Pipelines and Generative AI
- 3:04Overview of unstructured data: documents, images, audio, and more
- 9:00How AI models convert unstructured data into vectors for efficient search
- 12:30Vector databases: Why approximate nearest neighbor search is powerful
- 15:50Building scalable streaming pipelines with Apache NiFi and Kafka
- 19:40Using Flink for real-time data enrichment and analysis
- 24:00Demonstrating real-time AI use cases with models like Hugging Face and YOLO
- 29:30Integrating text, images, and other data sources into your AI workflows
- 33:17Practical implementation tips for building and running streaming AI systems
- 34:00Hands-on demo: Real-time processing with AI models and data pipelines
- 35:10Best practices for using APIs and external models for AI-driven pipelines
- 38:00Troubleshooting and debugging AI pipelines in real-time
related talks
53mApache Deep Learning 302
43mSentiment Analysis in Action: Building Your Real-time Pipeline
59mPolyglot Pipelines with Apache Nifi
52mMachine Learning Data Pipelines
50mThe tech behind modern AI/ML use cases in Gaming
31mBuilding Real-time Pipelines with FLaNK: A Case Study with Transit Data
46mEvent Streaming and its Potential for the GenAI World
45mReal-time, Streaming Advanced Analytics....