Double-Blind AI Evals with PySyft | Full 10 Min Walkthrough

In this walkthrough, you'll see how PySyft lets two parties evaluate a model without either side exposing confidential assets. What the demo covers: - Setting up a PySyft server locally - Submitting a model for evaluation without sharing weights - Running an eval dataset without revealing test cases - Reading results both parties can trust Try it yourself: https://github.com/OpenMined/PySyft **Timestamps:** 0:00 Overview 0:24 The parties 1:02 Spinning up the enclave 1:46 PySyft's peer-to-peer setup 2:43 Attesting the enclave 3:20 Benchmark owner 4:27 Model owner 6:21 Job submission + dual approval 7:40 Reading the obfuscated code 8:47 Running the final job 9:55 Results OpenMined is a non-profit foundation creating open-source technology infrastructure that helps researchers get answers from data without needing a copy or direct access. Our community of technologists is building Syft: the public network for non-public information. Learn more → www.openmined.org Supported by the OpenMined Foundation, the OpenMined Community is an online ecosystem of over 17,000 technologists, researchers, and industry professionals keen to unlock 1000x more data in every scientific field and industry. Join us on Slack → slack.openmined.org LinkedIn → www.linkedin.com/company/openmined X/Twitter → x.com/openminedorg Facebook → fb.com/openminedorg