Portfolio

See our portfolio of work.

Here are some projects that came out of the Data Science Accelerator. Read the plain-language case studies for a brief overview of each project. Read the technical papers for an in-depth description of the technical approach. We describe why each project is a good fit for the Data Science Accelerator.

Detecting cash benefit theft

The California Department of Social Services (CDSS) disburses cash benefits, through a welfare program called CalWORKs, to about 860,000 people every month. The disbursement mechanism, called Electronic Benefits Transfer (EBT), loads cash aid onto a debit card. Between June 2021 and March 2024, criminal organizations stole $194 million in from CalWORKs beneficiaries. We worked with CDSS to develop a predictive model to identify compromised cards and ultimately reduced food and cash benefit theft by 83%, saving the State of California millions of dollars. See the press release from the Governor's Office, the case study, and the technical paper.

Why is this project a good fit for the Data Science Accelerator? We build predictive models that enable proactive early detection.

Streamlining housing development

California has an ambitious goal of permitting the construction of 2.5 million homes by 2030 with 1 million homes being affordable for lower income levels. As a result of underproduction, housing costs have soared. Recent laws aim to alleviate the issue by building more affordable housing faster. We built a tool that allows California Department of Housing and Community Development (HCD) to streamline the permitting process. We identified projects that take longer to build than other, similar ones. We also calculated affordability metrics. We automatically verified 400,000 projects from the last 5 years that qualify for streamlining, saving HCD more than 2,000 hours of staff time annually. See the case study and the technical paper.

Why is this project a good fit for the Data Science Accelerator? We identify outliers in datasets to optimize resource distribution and reduce backlogs.

Forecasting community water system outages

The State Water Resources Control Boards' Division of Drinking Water (DDW) monitors 2,866 community water systems across California. In rare cases, and during periods of extreme drought, communities run out of water and the DDW funds deliveries of bottled or hauled water. With a few months of advance notice about which communities might face problems, the DDW could investigate and mitigate issues beforehand. We worked with DDW to build a predictive model that forecasts drought impacted water systems. Automatically running this easy-to-understand model every year saves DDW hours of staff time, avoids mistakes, and, ultimately, averts public health disasters. See the the case study and the technical paper.

Why is this project a good fit for the Data Science Accelerator? We build predictive models that enable proactive early detection.