The goal of this project was to test an enterprise AI model to see how it performs on advanced mathematical problems, nontrivial quantitative tasks, and multistep long-horizon tasks. 

The verifiers had to be deterministic, meaning they had to be rule-based with no use of AI. 

The deliverable was:

1. a suite of tasks in each category with their verifiers,
2. an infrastructure in Python that implements the verification with a dashboard that summarizes the results, and 
3. a report with recommendations.