Data Science in a Cloud World: What Every Data Scientist Needs to Know
summary
Company migrates over six years to a cloud-native stack and uses containers, Kubernetes, GPU instances and third‑party conversational agent AI services to ship a chat bot into TurboTax within months. Data pipeline distributes training across a 100-node cluster to build four million time series from a billion transactions, enabling personalized models at runtime. A 2 million‑transaction CNN experiment that takes three days on a laptop completes in one hour vs three days on scalable cloud GPUs, shifting focus to decisions per minute.
AI-written summary · about
chapters (9)
- 0:00Technical Vision Talks
- 6:11It takes massive compute to solve such complex problems
- 7:17Cloud adoption is mainstream and competition is fierce
- 7:27Every cloud provider offers a whole host of capabilities
- 8:04Al services are the differentiating factors
- 9:53One cloud doesn't have to fit all
- 13:22Distributed training is a game changer
- 18:04With the cloud, we've solved difficult customer problems
- 18:18Good luck on your journey!
related talks
31mA Journey of Migrating Data Platform from Data Center to Cloud
31mAccelerating AI with Hyperscaler Platforms
29mData Science in Production
01h 03mDiscovering the value of a cloud data platform
33mCloud & data - The Engine & Fuel of Business
44mPeople, Technology, Culture: How to make Data Science work
15mEnabling Scalable Data Analytics for Enterprise Success
19mLessons from Building a Large-scale, Multi-cloud Data Platform at Databricks