Stanford RAIN/OR (Recent advances in AI, INcentives, and Operations Research) is a seminar on the theory and practice of AI, incentives, and operations research. It serves as a hub for talks and discussion at the intersection of these fields and society. It is supported by Stanford’s Society & Algorithms Lab (SOAL), Stanford OpenLab, the Stanford Center for Computational Market Design, and the Stanford Computer Forum (and is open to Computer Forum affiliates).
Autumn 2026 talks are held in person on Tuesdays from 4:30–5:30 PM PT; room assignments are listed below.
Talk details
In this talk, I will describe a research program on time series that I have been pursuing with various collaborators over the past decade. I will start by recalling the literature on time series. I will discuss a representation for time series data through the lens of low-rank matrices and its implications for robust, scalable forecasting, imputation methods. Finally, I will touch upon ongoing project with the goal of developing efficient time series foundation model with synthetic data inspired by the traditional statistical models.
Bio: Devavrat Shah is an Andrew (1956) and Erna Viterbi Professor of Electrical Engineering and Computer Science at MIT. He is chief scientist at Celonis, Inc. He co-founded Celect in 2014 (part of Nike since 2019) and Ikigai Labs in 2019 (part of Celonis since 2026). At MIT, he was the faculty director of Deshpande Center for Tech Innovation between 2020 to 2023, founding director of the Statistics and Data Science Center at MIT between 2016 to 2020. He is a distinguished alumnus of his alma mater IIT Bombay. His work has been recognized through career prizes 2008 ACM Sigmetrics Rising Star, 2010 INFORMS Erlang Prize, 2024 INFORMS Applied Probability Society Markov Lecture, 2025 ACM Sigmetrics Achievement Award; paper prizes at IEEE Infocom, ACM Sigmetrics, NeurIPS, INFORMS Applied Probability Society, INFORM Management Science and Operations Management; INFORMS George B Dantzig thesis prize and test of time awards at ACM Sigmetrics. He is Kavli Fellow of National Academy of Sciences. He was an adjunct professor at Tata Institute of Fundamental Research, Mumbai and currently a distinguished professor of Data Science and AI at Indian Institute of Technology, Guwahati.
Keegan Harris
University of California, Berkeley
4:30–5:30 PM PT·Y2E2 299
Inference Auctions
Talk details
When inference demand exceeds available compute capacity, model providers must decide which requests should be served first. Users have different tolerances for delay from an LLM API, but current priority pricing schemes compress these differences into coarse fixed-price service tiers. We design an inference auction that allows users to bid for faster service. Our auction allocates priority in an economically efficient way without sacrificing latency, and we develop fast algorithms for implementing prices that incentivize truthful bidding. We also design an autobidding agent for our inference auction, where users specify an inference budget and the autobidder dynamically adjusts its bids over time to maximize user utility subject to the budget constraint. Experiments validate the practicality of our auction: it increases system welfare while maintaining the cache utilization and latency advantages of SGLang, a state-of-the-art inference serving framework.
