How to leverage the Apache Kafka Ecosystem to productionize Machine Learning
summary
Shows how to run trained models at scale by streaming sensor data from cars through an MQTT proxy into Kafka, replicating to an analytics cluster, and using Kafka Connect, Kafka Streams and KSQL for preprocessing, model deployment and real‑time inference. Demonstrates embedding H2O/TensorFlow models as a KSQL UDF to produce live anomaly scores, avoids RPC scale issues with model servers, and cites real deployments (LinkedIn ~4–5T messages/day, Netflix ~6PB/day) to prove it works.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:39Connecting the dots
- 1:38What is Machine Learning
- 2:33Machine Learning Algorithms
- 3:43Building analytic models
- 4:50Hidden technical depth in ML
- 6:13Examples Netflix Uber PayPal
- 7:27What is Kafka
- 9:24Kafka use cases
- 14:34Kafka infrastructure
- 17:17Preprocessing
- 18:12Case Sequel
- 21:39Data ingestion
- 25:01Model deployment
- 29:06Short examples
- 31:58Demo
- 38:24Conclusion
related talks
42mHow to Leverage the Apache Kafka Ecosystem to...
27mBeyond Logs: Real Time Anomaly Detection with Machine Learning
19mAI Workflows with Kafka and Modern Architectures
37mBig Data Analytics, Machine Learning, Real Time Processing
27mDeploying distributed AI and machine learning in financial services
01h 02mBig Data Analytics in Near-Real-Time with Apache Kafka Streams
20mShip your Machine Learning Application #bbuzz
52mMachine Learning Data Pipelines