In this use-case a pharmaceutical company has a model which can predict the occurence of a disease. They already had data from their clinical trials and are interested in having more of them coming from clinics in a different country. The clinics are naturally very cautious with their patient data and therefore are not willing to send the data outside their jurisdiction. At the same time, the clinics do not possess the required hardware to train the model locally.
Through avato the pharmaceutical company is able to:
1) Easier convince the clinics to participate in the research since no raw patient data will ever leave their premises;
2) they are able to securely link and train their model on the combined raw data without ever seeing it.
The parties connect to our platform (avato) through either a graphical interface or an API. The clinics encrypt their data locally and send it over to avato while the pharmaceutical company sends over the locally encrypted model. The model gets applied to the data and performance scores are computed on the predictions. Then avato encrypts the performance scores and sends them over to the parties. This process requires minimum tech expertise on the data owner side and provides zero overhead on the workflow of the data analyst. At all times it is guaranteed that nobody can access the data or model.