This August, we celebrated another successful year of the OpenDP Interns & Fellows program. We hosted three Visiting Fellows: Damien Aymon from the Swiss Federal Statistical Office; Shlomi Hod from Boston University; and Praneeth Vepakomma from MIT and three interns onsite in Boston: Tudor Cebere from Inria; Gurman Dhaliwal from UC San Diego; and Maxine Park from Harvard University.
The 2024 Visiting Fellows and Interns worked on several projects to help improve the OpenDP Library functionality, create more visualizations, and apply differentially private methods to real-world applications. Learn more about these projects from from three participants in the 2024 program below:
Damien Aymon (OpenDP Fellow)
Q: Tell us a little about yourself!
I work as a data engineer at the Data Science Competence Center (DSCC) of the Swiss administration, in the Swiss Federal Statistical Office (FSO). Our work at the DSCC is split between data science projects for other offices in the public sector and internal innovation projects. We are developing Lomas: a platform for remote differential privacy that includes the OpenDP library! In my free time, I am a passionate cyclist and mechanic.
Q: What are you currently working on?
I am working on a project aiming at making the disclosure risk more tangible for differentially private releases. Building on the concepts of f-differential privacy, we introduce the notion of relative risk: the highest possible knowledge gain of an adversary in the context of membership attacks. The project includes the additions of these tools into the OpenDP library to make them accessible to the widest possible audience.
Q: How is that impactful for OpenDP?
Budget selection for differentially private releases is far from trivial. Given a budget, the disclosure risk is difficult to interpret, which makes the privacy-utility trade-off a challenging problem to communicate. Adding tools and visualizations to the OpenDP library that help reason about this problem should make differential privacy more accessible and hopefully widen the community!
Q: What interested you in becoming an OpenDP Fellow?
While differential privacy is well accepted in the scientific community as the standard tool for controlling disclosure risk, it has yet to be widely adopted in practice for statistical releases by the public sector. The ongoing collaboration between OpenDP and the FSO has already sparked growing interest and demand for differential privacy within the FSO and beyond. I was eager to collaborate with the team at OpenDP to provide methods that make differential privacy easier to use and hopefully benefit the OpenDP community as well as the Swiss public sector.
Tudor Cebere (OpenDP Intern)
Q: Tell us a little about yourself!
I am a second-year PhD student at Inria, supervised by Aurélien Bellet. I am interested in differential privacy and how to create useful machine-learning models with strong privacy guarantees. Outside of work, I enjoy experimenting with new recipes in the kitchen, I enjoy spending time with my friends and am currently reading Avi Widgerson’s “Math and Computation”.
Q: What are you currently working on?
Differentially private algorithms require that the amount of information an individual can contribute is bounded, but in many real-world scenarios, there are no natural limits. I’m developing methods to determine ideal bounds for various scenarios, both in classical statistical problems and machine learning applications, both for classical statistical problems and for machine-learning related problems. Besides that, I enjoy writing Rust and I enjoy understanding the library’s low-level building block components and how to private them.
Q: How is that impactful for OpenDP?
Having a principled way of picking these bounds based on any priors data scientists have can improve the utility of many differentially private queries, like sums or means, providing better overall utility for the library.
Q: What are you planning to do after your internship?
I will enjoy a few days off in the US, and then I will return to France to continue my PhD and continue contributing to OpenDP.
Gurman Dhaliwal (OpenDP Intern)
Q: Tell us a little about yourself!
I recently graduated from the University of California, San Diego with a B.S. in Data Science. I’m also passionate about the intersection between tech and policy, with experiences in government, social science research, and nonprofit work promoting responsible data science and tech policy. In my free time, I love cafe hopping, running, and reading Greek mythology.
Q: What are you currently working on?
This summer, I’m working on projects aimed at making the OpenDP library more user-friendly for developers without a DP background. I’m enhancing the documentation for the Context API and the new Polars integration through tutorials that cover the data science workflow. Additionally, I’m conducting a data custodian stakeholder analysis to better understand the perspectives of the research and data communities on differential privacy and how we can engage with them.
Q: How is that impactful for OpenDP?
OpenDP has many powerful tools, but their functionality is often not easily accessible to programmers without a DP background. My work focuses on:
- Understanding current user needs and feedback.
- Creating comprehensive tutorials for new OpenDP features, helping users apply them to their projects.
- Developing a living document of common errors, their meanings, and solutions. As the first user outside the developer team to use some of these tools, I aim to identify and document potential issues new users might face, making OpenDP more approachable and reducing roadblocks.
Q: What are you planning to do after your internship?
I’ll be working as a Data Scientist in the investment management industry.
If you are interested in applying for the 2025 program, make sure to subscribe to our mailing list and check back on our website for announcements in October. In the meantime, you can reach out to Nina Wattenberg if you have any questions about the program. Congratulations again to our 2024 interns and fellows, and we look forward to welcoming a new cohort in 2025!