AWS SageMaker
Designing self-service ML infrastructure
Shared GPU infrastructure. Guided admin setup. A faster path from available compute to a working environment.
~99%
Setup reduction
From 1 to 2 days down to 2 to 5 minutes
- Role
- Lead Product Designer and Design Engineer
- Team
- Solo designer with engineering and product
- Duration
- 8 months
- Scope
- Shipped admin experience and data scientist prototype
- ~99%Setup reduction
- From 1 to 2 days down to 2 to 5 minutes
- 1:15Template reuse
- Templates could span namespaces and teams
- 3+Enterprise customers
- Adopted within 2 weeks of launch
- 13+Spaces created
- Active early usage after launch
Two workflows. One GPU cluster.
Project context
- Problem
- Admins needed governance. Data scientists needed available compute.
- My decision
- I shipped the admin foundation and prototyped a hardware-first path for data scientists.
- Tradeoff
- Data scientists launched with CLI access while the GUI remained a prototype.
- Shipped result
- Admin setup fell from days to minutes.
From manual setup to Quick Install
Install the foundation
Admin workflow. Dependencies and installation progress stay visible.
See capacity. Then create.
Find available compute
Data scientist prototype. Compare capacity by hardware or team.
Create a Space
Data scientist prototype. Choose hardware, then configure the environment.
A public milestone
