Last week, on February 6, Facebook launched a new call to invite the academic community to submit proposals for the Statistics for Improving Insights, Models, and Decisions request for proposals (RFP). The RFP is a joint effort led by the Facebook Infrastructure Data Science and Core Data Science (CDS) teams and builds upon the 2019 research awards in this area.

Data scientists at Facebook conduct global, quantitative research at scale to help improve the products used by billions of people every day. Infra Data Science is a team of applied quantitative and computational experts that use math, statistics, and machine learning to measure and optimize performance, reliability and efficiency of Facebook’s infrastructure and global telecom systems. “We are interested in leveraging recent advancements in statistics and machine learning to improve the performance, reliability, and efficiency of Facebook’s infrastructure,” says Rajiv Krishnamurthy, Research Data Science Director. “With this RFP, we hope to connect with academia on these topics and foster new collaborations.”

The Facebook CDS team is a research and development team working on improving Facebook’s processes, infrastructure, and products that enable more than 1.5 billion people to communicate with one another every day. A goal for CDS is to be on the forefront of innovation and partner with academia to develop state-of-the-art algorithms and methodologies. “I’ve been particularly excited by our work on measurement and evaluation under uncertainty because we can’t solve what we can’t measure,” says Aude Hofleitner, Facebook Manager and Data Scientist. “For that reason, we’re quite excited about receiving proposals in the areas of learning and evaluation under uncertainty, as well as efficient sampling and prevalence measurement.

“We’re also focusing on understanding how behavior and interaction can be indicative of malicious activity, and we’re looking forward to working with researchers who submit proposals in the areas of statistical models of complex social processes and anomaly detection,” Hofleitner continues. “We care deeply about providing transparency and control to the people using the Facebook family of apps, which makes work in the area of interpretability techniques for AI models a high priority for us.”

The CDS team also seeks proposals for causal inference and experimentation, an important topic for Facebook researchers on the Adaptive Experimentation team. Adaptive experimentation is one of Facebook’s tools for optimizing over large decision spaces. In 2019, Facebook released two tools for adaptive experimentation: BoTorch, a research framework for Bayesian optimization, and Ax, an experimental management system.

Ax and BoTorch enable anyone to solve challenging exploration problems in both research and production — without the need for large quantities of data. The CDS team discusses their open source work in the video below.

“Engineers and decision-makers routinely utilize experiments — the gold standard for decision-making — to make product improvements and optimize our infrastructure,” says Eytan Bakshy, Facebook CDS Research Manager. “We are excited to support any research that allows us to rapidly explore and understand large decision spaces, both in cases where we can and cannot run experiments.”

Bakshy is a key organizer of the Adaptive Experimentation Workshop, which is taking place at Facebook New York on February 13. The workshop aims to bring together distinguished academics and industry researchers to discuss the present and future for a research agenda around better decision-making through adaptive experimentation.

With a similar goal in mind, both the Facebook Infra Data Science and CDS teams seek to foster further innovation and strengthen collaborations between industry and academia through this new research award opportunity.

The deadline to submit research proposals is April 29, 2020, at 5:00 p.m. AOE. For more information and to apply, visit the application page. To view our currently open research awards and to subscribe to our email list, visit our Research Awards page.