Federal Grant Award Summary

Cornell University received a $451,594 CAREER (Faculty Early Career Development) Project Grant from the National Science Foundation's Division of Computing and Communication Foundations (CFDA 47.070: Computer and Information Science and Engineering) awarded April 15, 2026, with completion targeted for March 31, 2031. The project develops advanced algorithms for efficient artificial intelligence (AI) model training and inference scheduling using Gittins Index theory. On the training side, the research produces new acquisition functions for Bayesian optimization that intelligently explore the design space of potential AI models, reducing training time and energy consumption while achieving higher model quality. On the inference side, the project delivers Gittins Index-based task schedulers that optimize throughput and minimize tail latency in AI inference workloads, enabling systems to process more tasks per second with reduced wait times.

Beyond AI applications, the research produces algorithmic tools with broader applicability to engineering design problems such as drug discovery and reactor design, as well as practical optimization solutions for non-computing domains including food delivery logistics and urgent care clinic scheduling. The research outcomes include theoretical descriptions, practical implementations, and validation studies demonstrating how the Gittins Index framework addresses optimization challenges in both the exploration phase of AI model development and the scheduling phase of AI system deployment. This work advances U.S. strategic capabilities in AI while contributing to reduced data center energy consumption in artificial intelligence systems.