Many companies develop proprietary machine learning models and provide access to it as a service. Often, the predictions have to be made on sensitive data that their clients are not willing to easily share. The most prominent example is the multitude of startups with a Machine Learning product, especially in the healthcare space. In most cases their clients have to upload sensitive information to company servers in order to get their prediction. This creates two main issues for the startups: 1) Getting traction is much harder because of this obstacle; 2) even when they get the data, they have to spend a lot of resources on keeping it safe.
Through our platform these companies can ensure their customers that no raw data is ever sent to the company while still providing their service with no extra hassle. On top of that, the companies don’t have to deal with securing their data against a breach because they never get hold of it.
In this case our platform is sitting between the model and the user. The company and the users encrypt their model and their data respectively and send the encrypted data to the cloud. There, our platform applies the model to the data and the encrypted results are sent back only to the users. The whole process is adding zero friction to both parties by being able to integrate completely with the existing workflow of submitting data to a public endpoint.