AI audit and certification

Algorithms are used to make important decisions. Because of the scale and speed of algorithms, issues with bias, fairness and accuracy can be greatly amplified leading to losses, impact on the company’s reputation etc

To mitigate this risk, Yields.io offers algorithmic audit and certification services. One of our unique components is that we leverage our award winning platform Chiron for automated model risk management. The automation allows us to perform an exhaustive in-depth audit, that can moreover be repeated whenever new data becomes available. This type of monitoring capability is necessary since AI applications are being changed and re-trained at a high frequency.

A typical algorithmic audit consists of the following steps:

• Week 1: model onboarding (we review the existing documentation and available data)
• Week 2-4: model audit: we assess the trustworthiness of the algorithms based on an best-practice approach inspired by ALTAI
• Week 5: Review of the draft audit report with the client
• Week 6: Final report and presentation of findings

After the algorithmic audit, you have the opportunity to subscribe to our algorithm monitoring service to receive periodic updates on the performance of your models.