ML anywhere: Single pane of glass for cross-Region & on-premises AI/ML (PRT066)
summary
Domino runs a central control plane in Frankfurt that manages code, access and config while lightweight Nexus data plane instances run locally (South Korea on S3, US on-prem Nvidia DGX, Italy behind VPN). Users view the same UI (JupyterLab, MATLAB, SAS) and click to ship workloads to a chosen region; clusters are EKS with Helm, containers scale to 120-core/GPU machines, models deploy to SageMaker, and data access is enforced so local datasets never leave their region — it works as described.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:43Domino Data Lab in 60 seconds
- 2:10Why hybrid cross-Region
- 5:03What Natalya needs to do
- 6:06Worldwide collaboration with Domino Nexus
- 9:31Share the code without sharing the data
- 13:45Summary
related talks
44mScaling ML from 0 to millions of users
29mML in Production on Databricks
34mScaling AI Like a Pro-PepsiCo’s LLM Deployment Strategy on K...
37mUniversal AI: Execute Your Models Where Your Data (And Users) Are
22mAI @Scale
44mPractical deep dive into machine learning techniques and MLOps
27mML Platform Journey
20mSecure, private LLMs in your cloud with Anyscale Endpoints