Anand Pradhan, ICE | Databricks Data + AI Summit 2024
summary
Built and running is a low-latency trading and market-data stack using InfiniBand, a hybrid cloud surveillance pipeline that mines 650 billion transactions per day at microsecond granularity, and a GPU-run BERT-based small language model for news sentiment. Workflows include time-series deep learning, vector databases for embeddings, fine-tuning open-source models before investing in a frontier model, and an internal AI Center of Excellence that enforces data governance, audit trails, observability, accuracy and explainability while forbidding LLMs for core pricing decisions.
AI-written summary · about
chapters (5)
- 0:00From trading tech to AI
- 10:34AI center of excellence
- 12:57Building versus buying models
- 15:32Data privacy and control
- 17:14Real-world AI application
related talks
35mHigh-performance data foundation for AI success
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases
49mGenAI for Business on the Modern Data Stack
16mTransforming Analytics With AI: A Comprehensive Guide to an Advanced Tech Stack
50mData Science in the Enterprise
43mArchitecting for Speed & Scale
01h 17mData Engineering in the AI Era: Building a Robust Foundation for AI Applications
52mElevate your data and AI governance with Databricks Data Intelligence Platform