Shaping the future of data science

Image: WiDS Cambridge Industry Panel. l-r: Fotini Christia, IDSS Director; Lisa Bonalle, ​Chief Executive Officer, GDS Link; Tina Reich,​ Independent Board Director and Advisor; Lisa Schirf, Managing Director, Global Head of Data & Analytics, Tradeweb; Julia Taitz, Data Scientist at Instagram, Meta; Vanessa Liu, Co-Founder and CEO, Sugarwork

For nearly ten years, the Women in Data Science (WiDS) Cambridge conference has been building a community of practitioners and showcasing cutting-edge analysis and computing.

Scott Murray | MIT IDSS

The Women in Data Science initiative, known as WiDS Worldwide, was first launched as a one-day technical conference at Stanford in 2015. The following year, WiDS ambassadors launched satellite conferences all over the world — including Cambridge, Massachusetts. Since its inception, WiDS Cambridge has been championed at MIT by the Institute for Data, Systems, and Society (IDSS).

“WiDS plays an important part in IDSS’s role as a connector,” said Fotini Christia, Ford International Professor of the Social Sciences in the department of Political Science and Director of IDSS. “In addition to networking and learning opportunities, we’re hoping to foster conversations that help data scientists find the challenges that interest them, the areas where they can make an impact, and to understand how they can collaborate.”

Data science applications from many different domains have been featured at WiDS Cambridge, which is co-organized and hosted by IDSS, the John A. Paulson School of Engineering and Applied Sciences at Harvard, and Microsoft Research New England. From human genomics to outer space, through deep dives on machine learning tools to panel talks on policy, WiDS creates a space for anyone who is interested, regardless of experience level, to learn from women doing exceptional work in the field.

“The representation matters,” said Shreyaa Raghavan, a student in the IDSS Social and Engineering Systems doctoral program who presented a poster on her research at WiDS. “It’s important to see yourself in the work. Having women advisors and mentors helped me to visualize myself in the field.”

While featuring women speakers, the conference is open to all to attend and focused on promoting data science skills and forming productive relationships throughout the broader practitioner community. “WiDS creates a great environment for making connections and answering questions,” said Raghavan.

Christia moderated a panel on industry perspectives and career journeys, and was joined by CEOs, board directors, and data scientists. Panelists discussed the importance of technical competence and talking the talk, but also how essential communicating goals, outcomes, and impacts are to success.

Executive Director of WiDS Worldwide Chisoo Lyons, who attended WiDS Cambridge in person, emphasized the value of this experience and leadership, and of the WiDS inclusive community, in her opening remarks. “WiDS is about shaping the future of data science and AI,” she said, “not just for women, but for everyone.”

From computation to carbon emission

Sustainability was the major theme of WiDS Cambridge 2025, with two panel talks on Sustainable Computing and a keynote on Efficient Computing for AI and Robotics delivered by Vivienne Sze, a professor in MIT’s Department of Electrical Engineering and Computer Science (EECS). Sze outlined how the energy consumption of computing has grown rapidly and by orders of magnitude due to emerging technologies from AI to big data centers.

“Deep neural networks deliver high accuracy in applications like autonomous vehicles and image recognition,” explained Sze. “This comes at a high compute cost.” The high energy consumption of these applications can translate into increased carbon emissions. Sze’s team works to develop more energy efficient computational methods. “Moving data costs energy. One solution is to exploit data reuse and reduce memory access to reduce data movement,” Sze shared.

Other approaches to reducing the energy cost of computation include embedding computation in memory to reduce the amount of data movement, and even computing with light to reduce the cost of data movement.

“The cost of moving a photon can be independent of distance,” said Sze. “So light can be relatively cheap to move around.”

Though energy costs remain a concern, AI tools can also bring critical new insights into research in climate and ecology. Panelists shared examples of this impact, from improving ocean models to account for small-scale physics to data methods in atmospheric science on annual wildfires.

MIT EECS professor Sara Beery also shared perspectives from her work in biodiversity, where researchers have ambitious goals of tracking changes at huge scales — across species, and across planet Earth.

“One of the challenges in biodiversity research is that we are never measuring directly, only using proxy data,” explained Beery. “In order to model and interpret this data, we need AI tools to translate raw data like trail camera images to actionable scientific observations.”

Measuring trade-offs in research impact versus energy consumption can be a daunting notion, but Beery also pointed out that not all AI is the same. “There can be lost nuance here,” she warned. “There are big differences in cost between, say, edge computing small-scale data and cloud computing. Not all data science is equally expensive.”

Building the systems we want to see

The afternoon panel at WiDS Cambridge 2025 approached sustainability in systems, exploring how AI can advance the design of future infrastructure systems. Such systems, such as power grids and communication networks, are difficult to model due to scale and complexity, and changes to them are incremental and slow.

“It’s safer to build something that is like what we had before because there’s data on that,” pointed out Civil and Environmental Engineering and IDSS professor Cathy Wu, whose research uses machine learning for control and optimization in transportation and mobility.

With new and emerging tools, however, bigger changes to efficiency are becoming possible, with impacts that extend beyond cost savings. “Traffic accidents are still a leading cause of preventable death in young people,” said Wu. “Highly accurate traffic behavior models are currently used internally by autonomous vehicles. If we bring that tech into infrastructure, we can deploy stop lights that can watch traffic in real time and make adjustments that reduce the likelihood of a crash.”

Wu’s research team includes Raghavan, who has analyzed simulated data modeled after traffic demand profiles, as well as data replicating real traffic settings such as bottlenecks. “Our goal is to figure out how much congestion we can prevent and control on real highway networks,” said Raghavan, who points out that decreased congestion also means decreased emissions. “Using real traffic data, we can find recurrent problems and direct infrastructure changes where they are most effective.”

“This work is about building the systems that we want to see in the world,” added Wu.

Building new infrastructure systems is an interdisciplinary challenge, as is developing new AI tools for specific applications. One bottleneck, according to Beery, is capacity: there simply aren’t enough people with domain specific knowledge and proficiency using data analysis and machine learning.

“Not everyone needs to be a machine learning researcher, but we need robust applications of machine learning to account for limitations, bias, and variability in data,” said Beery, who trains ecologists in machine learning techniques while teaching data practitioners about biodiversity.

Growing and training the data science community is what WiDS is all about — and it’s core to the IDSS mission, as well.

“IDSS’s academic programs and online courses produce data literate problem-solvers,” said Christia, “and WiDS is one way that we are making sure there is a robust network of opportunities and mentorship for our students, our online learners, and really for anybody who wants to tackle big challenges in data science.”

WiDS Cambridge will host a datathon workshop in April consisting of a data science and machine learning tutorial, followed by a team-based practical session focused on a single data science task. Participants of the workshop will participate in the Global WiDS Kaggle competition focused on studying women’s brain health.