Using Kaskada to do Real Time Machine Learning
summary
Cascada is a real-time timeline engine with a Python API and a Rust core that stores processing in Apache Arrow and can use GPU, offering native time travel, temporal joins and windowed features for online ML; benchmarks claim consistent speedups over DuckDB and an order-of-magnitude improvement versus a 63-page Spark churn pipeline reduced to two Cascada pages. A live demo builds a RAG chatbot against Astra DB using text-embedding-ada vector search to return three nearest rows in seconds and a Slack notifier called beepGPT on GitHub.
AI-written summary · about
chapters (10)
- 0:00Introduction and speaker background
- 2:33What is Kaskada and its purpose
- 10:49Real-time AI with Kaskada features
- 11:10Building a RAG app in 5 minutes
- 13:25Demo setup and dependencies
- 18:56Vector search and embeddings
- 23:31Combining real-time AI and generative AI
- 26:24Fine-tuning and Slack bot example
- 28:54Production challenges and future
- 31:26Open source and closing remarks
related talks
41mUsing Kaskada to Do Real Time AI- 41mUsing Kaskada to Do Real Time AI
40mOvercoming data infrastructure limitations with a new paradigm for ML
19mBuild Real time, AI powered Apps
36mMoving Faster: Why Intent Media Chose Cascalog
39mAccelerating Data Science with RAPIDS
45mManaging Data & Analytics Projects Effectively
30mArchitecting Applications With Multiple Open-Source Big Data Technologies