Contributor
arpitvaghela

Integrating Pandas into Syft


Mentors
Tudor Cebere, Madhava Jay
Organization
OpenMined

The Syft ecosystem allows one to write software that can compute over information you do not own on machines you do not have (total) control over. PySyft mainly allows direct computation on unseen data and generates static graphs of computation that can be deployed or scaled later.

To do so PySyft creates an AST that maps function calls to their exact path and knows what to do with a node in the tree. AST allows remote execution, and to do so AST provides a local handler for the result of remote execution through pointer alongside generating all the permitted methods and attributes on a Pointer.

The current implementation has integration for pandas’ DataFrame and Series, i.e, one can remotely execute methods and access attributes on objects on these classes using their pointers. This is to be further extended for all the classes and global functions in pandas.