The Challenges of Deploying Real-time AI for Finance and how Open Source can help
summary
Real-time finance ML needs millisecond P99 latency, not nightly batches, and companies deploy open source Feast and Feathr with Redis as the online store to hit 50–200 ms windows. Benchmarks cited show Redis delivering much higher throughput and being 14x cheaper than DynamoDB in one comparison; Cash App cuts tail latency from ~30 ms to ~10 ms after moving to Redis. Large feature stores reach terabytes and can drive >50% of MLOps cost, so Redis Enterprise and Redis on Flash are used to lower cost while keeping sub-100 ms performance.
AI-written summary · about
chapters (10)
- 0:00Start
- 4:38Leveraging Open Source for Feature Stores for Real-time Al/ML
- 7:10Open Source the Driving Force for Innovation for AI/ML
- 10:02What are feature stores?
- 16:19Data pipeline for real-time Al/ML
- 20:13Leveraging Redis data types and structures for feature stores
- 24:49Feature Store: The Cornerstone of MLOps Platform
- 31:22Companies leveraging Redis and Redis Enterprise to successfully deploy Online Feature Stores for real time Al/ML
- 39:52Feast on Azure with Redis
- 41:14The most beneficial areas for Open Source engagement in Finance
related talks
19mThe Key Role Of Feature Stores In Deploying Real-time AI
19mML feature store a cornerstone in modern data architecture for AI/ML
19mBuild Real time, AI powered Apps
29mBuilding Resilient AI Platforms: Financial Fraud Detection
38mRealtime Data Activation - Analytics, Intelligence & Decision Making
16mOperationalizing AI in Finance: Real-Time Analytics
34mUltra-Reliable Low-Latency Solutions for Industry 4.0 Predictive Maintenance
24mUnder Pressure: Applying ML in Real Time