In this sprint our Data Science team created a demo, that shows how you can do so called “data centric federated learning” on data that lives on different servers, making that data accessible in privacy preserving manner to data scientist over a private network. An important aspect of federated learning is the execution and download permissions. We added code to pysyft, making it easy to pre-supply your node with permissions around sets of operations on your data. Apart from the infrastructural side of the demo, we focused on the machine learning aspects, implementing the first transformers in pysyft. Also much effort was put into helping with reviewing Memri plugins, improving the acceptance criteria for plugins, and answering questions about using pymemri.
Within Platform we are quickly progressing to our goals - creating alpha product. Our iOS team was busy implementing syncing with POD and our Flutter team improving UX and creating Detail View with Timeline for Person with Subviews (which will provide ability to create more interesting CVUs). Have a look at our new view for Person.
A good handoff is no cakewalk, and during the last sprint our product design team was focused on smoothing out the wrinkles in the process to align the design and development flow with the needs of rest of the team.
Within the product, we are constantly optimizing the on-boarding experience of the app to prepare the most promising variants for further testing, updating features according to the quarterly plan, and last but not least, experimenting with UX patterns and flows for future releases.
Also we've hired and on-boarded a new member joining the Memri-team:
Muhammad Bilal recently joined Memri for our DevOps role. He'll be working in Platform ME to create a Hosted Memri/Pod/Plugins infrastructure and help us with building pipelines, team access policies and other things traditionally belonging to DevOps / System Administrators. Welcome aboard!