From Quick Wins to Revolutionising Productivity & CX with GenAI
summary
Claims a pipeline that vectorize entire videos with timestamps, stores vectors in DataStax Astra DB, and answers natural language queries by jumping the UI to precise moments. Implementation uses ViT‑B32 for frames, Whisper for audio chunking, and Langlow to wire embeddings, prompts and a REST endpoint. Demo retrieves medal moments reliably but reveals limits: small image model, chunking granularity and start vs end timestamps need tuning, hybrid search across multiple videos and local open‑source models remain practical and cost effective.
AI-written summary · about
chapters (10)
- 0:00GenAI and semantic search
- 3:17Data Stacks platform overview
- 6:52Multimodal apps and vector search
- 9:17Demo setup: video/audio
- 12:34Video frames and embeddings
- 17:26Audio processing challenges
- 22:24Building Langflow AI
- 25:35Live search demo
- 27:43Tech stack and deployment
- 28:27Q&A
related talks
57mEnter The Brave New World of GenAI with Vector Search
33mVector Databases for GenAI applications
19mPOC to PROD: Hard Lessons from 200+ Enterprise GenAI Deployments
12mCodeless Generative AI Pipelines: Prompt Automation
32mBuilding a ChatGPT Data Pipeline with RisingWave Stream Processor and Cassandra Vector Search
18m: Workforce Efficiency with AI-Driven Vector Search
33mArchitecting Scalable Multi-Modal Video Search
29mEnhancing Vector Databases with Real-time Stream Processing