Our 2023 Fellows Reflect on Their First Six Months at xD

I’m working on the Database Modernization project that is focused on moving on-premises servers into the cloud. This is a multi-year project that will assist the Census Bureau with open-source technology adoption. Additionally, I am participating in this year’s Data Science Training Program with a Machine Learning Specialization. My goal with being in this program is to help me grow in foundational and application-specific skills for model creation and assessments.

The biggest learning I had from my project so far is being open to change. Over time, the focus of my project has shifted to higher priority data migrations. So being flexible and continuously seeking the root problem that we are solving has been key.

My favorite thing about xD is everyone in this cohort. Every person brings such a unique skillset to our group. This makes it super easy to learn new tools and technologies from each other. Moreover, each person's understanding of when to lead and when to follow makes this such a collaborative environment. I can’t think of better people to share this experience with.

• Successfully develop a machine learning model that will add value across the Bureau

• Complete the development of the MVP for the IRE-EDL data migration

My work right now focuses on how to use privacy-enhancing technologies (PETs) to connect sensitive datasets that have different owners in a secure way. The long-term goal of that work is to enable third-party audits of machine learning models so that engineers and practitioners who build these models can detect disparities or biases in their work both reliably and quickly.

One of the biggest things I learned in my first six months at xD is that even though starting out a new project can be a bit overwhelming at first, the key is to ask for help when you need it and not get too stuck in the mud trying to figure things out yourself. It’s another reason that I so appreciate the awesome team that we have!

One of my favorite things about xD is the stellar team of people we’ve put together. Everyone on the team is both extremely friendly and extremely good at what they do.

• Create a demo to show how we can use PETs to enable third party audits of ML systems

• Work towards using PETs to join datasets across different federal agencies

• Publish my work in blog posts and papers

I’m working on user-centered model cards for transparent documentation of government applications of machine learning and a privacy-enhancing technologies lab to help policymakers and developers implement best practices for sharing sensitive information while protecting privacy.

Thinking about technical challenges with social impact can be daunting. Developing creative solutions together has helped build nuance into our team’s approach. I’ve learned so much through interrogating gaps in my own understanding and workshopping ideas with other fellows.

The most exciting thing about xD is the opportunity for collaboration within our awesome cohort of fellows, across government, nonprofits, academic labs, and industry.

• Publish work on model cards and ML audits using privacy-enhancing technologies

• Onboard the next emerging technology fellowship cohort!

I am currently working on a pilot project to establish a secure multi-party computation between Census and the Bureau of Economic Analysis.

I’m learning more about surveys and data at the Census Bureau. I didn’t know that the Bureau conducts more than 130 surveys each year. I’m also learning about Title 13 which protects the confidentiality of data and survey respondents.

In addition to great colleagues, I love the forward-thinking and experimentation-oriented culture of the team.

• Facilitate an MVP for the Census-Bureau of Economic Analysis SMPC project

• Present at one of Census’ affinity groups

I’m working on privacy-enhancing technologies (or PETs) both with a group at the United Nations and with NIST, the National Institute of Standards and Technology. We are working on deploying, testing, and documenting PETs for different use cases. Another area I’m contributing to is bias mitigation in AI applications. Specifically, I’m creating proofs-of-concept and tools for Census Bureau teams such as a model card generator. I’ve also been helping to maintain and update our team website.

Keeping stakeholders engaged is key to moving a project forward and along the correct path. Also, working in government often requires the review or approval of several different parties, so it’s important to have different projects you can pivot to during times when you’re waiting on work from others.

The best part of being on the xD team is being able to work with a great group of people with diverse skillsets and backgrounds on lots of interesting applications in government.

• Publish a paper and give a conference presentation about our ongoing privacy-enhancing technology projects

• Add new sections and features to xd.gov to increase our web presence and better publicize our work

• Hire and onboard new fellows!