Training data is often biased by the world long history of human undue prejudices. If an algorithm is trained on a biased training set, it will simply codify and amplify this bias. That's why we built a set of auditing algorithms to highlight and remove any potential biases from the company past data.

Algorithms are not objective and fair by default. That’s why we built inclusive models that ensure high standard of fairness—such as allocation, representation, gender, ethnicity, stereotypes and more.

Using online learning, a process which enables a computer to learn continuously, our models get smarter with every interaction and feedback.