Managing decisions with Amazon Kinesis for better outcomes
summary
Combines Kafka/Kinesis/MSK ingestion with a lightweight Go-based flow runtime that claims 10–50× lighter than JVM apps, offers stream ops (joins, aggregates, tumbling/sliding windows) and edge deployment, layers a contextual declarative rules engine that keeps state to disambiguate events, and runs TensorFlow inference. Introduces Catalyst ML transforms (JSON spec) to automate the data-prep work that takes ~80% of ML effort, shows a live fraud-detection pipeline: preprocess streams, apply ML model, and act in real time; available open source and as an AWS service.
AI-written summary · about
chapters (6)
- 0:00Introduction to decisions
- 1:44Real-time decision models
- 4:08Project Logo overview
- 7:07Streams and rules engines
- 9:47ML and data transformation
- 11:42Demonstration and Q&A
related talks
20mBatching vs. Streaming - Scale & Process Millions of Measurements a Second
47mBuilding High Performance Real-time Stream Processing Microservices
36mStream All The Things!
47mStream All the Things!
14mPerformance Optimization in High-Volume Transaction Systems
39mDeploying Batch and Streaming Architectures on AWS
28mTransforming the Financial Services Industry with Event Stream Data
24mUnder Pressure: Applying ML in Real Time