by OpenDP Team
Reposted from the OpenDP blog. You can see the original blog post here.
The 2023 OpenDP Community Meeting was held after the TPDP conference on Friday, September 29th at Harvard University. Kicking off the meeting, we held a welcome reception at the Science & Engineering Complex outdoor terrace where over seventy attendees got the opportunity to meet and start networking on the previous evening. Attendees enjoyed food and drinks with beautiful weather while chatting in advance of the meeting the next day.
Community Meeting Highlights
The OpenDP Community Meeting, held at Harvard’s CGIS South building, hosted over 150 in-person attendees with approximately thirty people joining us live online. This was the first year we had an in-person component and it was a thrill to see so many people coming together and talking about a broad range of subjects around differential privacy and the OpenDP initiative.
Encouraging an inclusive and active OpenDP community was one of the main goals of the Community Meeting. We were keen to ensure the agenda included a variety of topics that included specific OpenDP updates but also presentations around differential privacy use cases, opportunities from community members via ten pitch anything talks, and two breakout sessions on key challenges may people face when dealing with data privacy. Recordings from the full agenda can be found on the home page of the OpenDP Community Meeting website or on the OpenDP Youtube channel.
Hackathon and Tutorial Highlights
On Saturday September 30th, the OpenDP team hosted a full-day session at the Science & Engineering Complex where twenty in-person attendees and twelve virtual participants got a hand-on tutorial working with the library tools and software. It was an interactive session that allowed for questions and suggestions by the group to be answered in real-time. Led by Andy Vyrros and Mike Shoemate from the OpenDP team, attendees were then invited to participate in a hackathon-style workshop. The stakes: bowling trophies for three categories: Most Creative Team, Most Useful Project, and Jury’s Prize.
The competition broke down into six teams spanning a broad range of topics: bounds estimation, graph-DP, the matrix mechanism, DP decision trees, a tighter conversion between privacy definitions, and DP for text data. The workshop had a non-competitive nature, with an emphasis on open collaboration.
In the end, the margins were razor-thin! “closeDP” (Jack Fitzsimons, Khadija Hafeez, Hardik Kapoor, Yanis Vandecasteele) won the Most Creative category for implementing DP decision trees and applying them to estimate causal treatment effects in randomized clinical trials. “Approximately Concentrated” (Felipe Gomez) won the Most Useful category for deriving/implementing a tighter conversion of zCDP guarantees to approximate-DP. We had a three-way tie between the “Dumb Dragons”, “Humann Arbitrager”, and Aishwarya for the Jury’s Prize!
You can find the OpenDP tutorial video for future reference on the 2023 Community Meeting website’s Tutorial and Hackathon page or on the OpenDP website.
Thank You and Next Steps
The meeting could not have been as engaging if it weren’t for our esteemed presenters, panelists, and pitch-anything community members! THANK YOU for taking the time to prepare and participate in this year’s agenda. To our sponsors – CapitalOne, Oblivious, Meta, and Knexus Research – we are grateful for your support in providing an enjoyable experience for all our attendees.
Keep an eye out for opportunities to continue the conversation, join working groups, and participate in workshops and activities! Make sure you’re on the OpenDP mailing list and in our slack channels to stay in the know!
Interested in learning more about the speakers or the panelists? Find the full list of speaker bios here and a quick Q&A with the industry panelists in this blog!
