The Harvard Data Science Initiative (HDSI) in conjunction with Harvard University Information Technology (HUIT) and the Office for the Vice Provost for Research (OVPR) has selected two projects for support through the Innovation API Fund, providing OpenAI API credits to accelerate research efforts across Harvard University. Learn more about the fund and how to apply on the Innovation API Fund page.
SW•Efficiency: Evaluating LLM Capabilities in Code Performance Engineering
Vijay Reddi, Professor of Electrical Engineering and Computer Science at Harvard University, spearheads a groundbreaking effort to evaluate how large language models can accelerate real-world code performance. Through SW•Efficiency, a benchmark of more than 500 optimization challenges drawn from widely used data science and machine learning libraries such as Pandas, NumPy, and Scikit-learn, the project investigates whether AI systems can autonomously detect performance bottlenecks, refactor code, and deliver measurable speedups while maintaining correctness. Using OpenAI’s GPT-5 and related models, the team will produce open-source benchmarks and reproducible results that advance the science of AI-driven performance engineering and redefine how developers approach software optimization .
Developing an LLM-Assisted Pipeline to Monitor Digital Well-Being Design Features in Social Media Platform Announcements
Led by Kasisomayajula (Vish) Viswanath, Lee Kum Kee Professor of Health Communication at the Harvard T.H. Chan School of Public Health, this project is creating a large language model–assisted pipeline to systematically review and evaluate social media platforms’ digital well-being design features. By analyzing more than 13,000 platform announcements across seven major companies, the team will generate new insights into how design changes affect user well-being and inform evidence-based policy recommendations. The work will culminate in an open-source tool and dashboard to monitor digital well-being practices at scale, advancing both public health research and platform accountability .