Purple

French expert platform for AI

Human expertise that trains and evaluates sovereign AI

Purple brings together experts in banking, insurance, law and finance to design evaluation rubrics, grade model outputs and test AI agents, on a fully sovereign infrastructure.

Three ways to work with Purple

You are an AI lab

Measure your models on Purple Bench FR, then let French-speaking experts produce your training and evaluation data.

Discover Purple Bench FR →

You are deploying an AI agent

Have your model or agent evaluated by professionals from your industry before going to production.

Evaluate my agent →

You are a domain expert

Put your experience in banking, insurance, law or finance to work for the best AI models.

See assignments →

Sovereign by design

SecNumCloud-qualified hosting targeted, mandatory two-factor authentication, no client data leaves its environment.

Learn more →

Prototype status

Application
Up
Database
Checking…