AI Compass: helping teams find the right foundation model
The number of AI foundation models keeps growing. AI Compass helps customers choose a "custom fit" model for their project and suggests the hardware best suited to run it.
Problem
There is an overflow of AI foundation models, and it is constantly growing. Teams evaluating them have no straightforward way to tell which model actually fits their use case, or what hardware they would need to run it well.
Process
- 1
Framed the problem with the Innovation Lab team using Lean UX: assumptions first, then the smallest thing worth testing.
- 2
Sketched lo-fi screens for model discovery and comparison, including a sessions model so exploration could be saved and revisited.
- 3
Ran user research with internal and customer-facing teams to understand how models get shortlisted in practice.
- 4
Built prototypes of the compare-and-recommend flow and refined them against research feedback.
Outcome
A working concept for model discovery — find your model, compare candidates, and get a hardware recommendation alongside the choice.