Live Q&A Session #1
summary
Composable data tooling is an extensible engine-agnostic library that runs inside different compute engines rather than replacing them, with migrations varying from trivial to multi‑month projects depending on engine maturity. Arctic reports training on 3.5 trillion tokens with roughly half processed in Snowflake (order of terabytes), using Snowflake for orchestration, annotation and rapid iterate‑evaluate loops. Production features include Razorpay’s magic checkout and LLM‑driven devops debugging and one‑click API migration, and incident workflows emphasize transparent confidence scores and human verification.
AI-written summary · about
chapters (17)
- 0:00Live Q&A
- 0:25Composable data systems growth
- 2:32Architecture design at product-market fit
- 4:20AI transparency in incidents
- 5:46Arctic data processing future
- 7:11Velox execution engine
- 8:13Razorpay merchant data privacy
- 9:59AI-assisted investigations
- 11:04Arctic multimodal data
- 12:08Migrating to Velox
- 14:55Razorpay GenAI innovations
- 16:20Human vs AI in incidents
- 17:43Pre-processing large data
- 19:49Composable data stack components
- 22:16GenAI operations at scale
- 23:44AI-driven incident scaling
- 25:09Arctic data scale
related talks
31mAI-driven DataOps: Building data products with a taste of the future
32mDataEngBytes 2023 - BRI-T1-10 Rishu Saxena & John Cosgrove
28mUsing Dataops To Build And Manage Data Products On Snowflake
27mAgile Data Platforms : Innovate at the speed of thought
25mPerfecting Our Modern Data Stack Recipe: What We Learned On The Journey At Huel
39mWhat Data Engineering Can Learn from Front-end Engineering
30mBuilding Data AI Application with the Snowflake AI Data Cloud
29mBuilding End to End Gen AI Powered Analytics: From Data to Production