PySyft

Why we included this project

Most analysis assumes you can copy the data you need, which is exactly what the people holding sensitive records won't allow. PySyft flips that assumption: instead of sending data to the analysis, it sends the analysis to the data. A data scientist writes a normal Python script and submits it as a job that runs inside the data owner's sandboxed environment, starting against mock data before touching the real records, and only approved results come back. Jobs travel through file-based channels like Google Drive that organizations already run, so there's no new infrastructure to deploy. For teams working with records they can't copy or share, that's a practical path to doing real research without ever negotiating raw access.