Building a Real-Time NBA Analytics App
summary
A real time NBA analytics app streams a clip of Game 7 2016 Warriors vs Cavaliers, sends video frames every second into Groq realtime LLM which emits live commentary and text embeddings, pushes those embeddings and metadata into SingleStore vector + full-text + hybrid search to power a live LLM chat feed and dashboards, uses an OCR model to pull the on-screen score and compute a live (advisory) win probability, reports per-response latency and counts, claims millions of events/sec, petabyte analytics, and the demo runs live (no audio).
AI-written summary · about
chapters (7)
- 0:00Introduction to speakers
- 0:22Why real-time AI applications
- 2:08Need for multimodal databases
- 3:56Problems with complex AI stacks
- 5:44Why SingleStore
- 7:32Real-time NBA analytics demo
- 10:40Demo features and wrap-up
related talks
27mStream, Store, Visualize: Fast-Track Real-Time Analytics the OSS Way
38mRealtime Data Activation - Analytics, Intelligence & Decision Making
51mVideo Analytics for Football games
28mMoving Towards Real-Time OLAP - With First Principle Thinking and Guided by Access Patterns
45mManaging Data & Analytics Projects Effectively
20mAI-Driven Real-Time Ball Tracking for Live Sports Streaming
28mTheory to practice: A guide to real-time architectures
41mAdding Generative AI to Real Time Streaming Pipelines