Hi! I'm an assistant professor of Operations Research and Information
Engineering (ORIE) at Cornell Tech and Cornell University, and an ORIE, Computer Science, and Information Science field member.
Our work: "Full Stack, Public Interest AI"
I advance high-stakes, public interest AI in the wild. My work connects computational systems, individual decisions, and societal outcomes, considering three interrelated roles of algorithms: (1) algorithms allocate societal resources and attention; (2) algorithms support individuals navigating these systems; and (3) AI helps researchers understand these systems. We consider these roles in various domains: Human-Algorithmic Market
Design, Foundations and Systemic Effects of Modern AI, and Operations and Algorithms for Government
My work is “full stack,” spanning the research-to-deployment pipeline: formulating mathematical models; conducting empirical analyses; developing statistical, optimization, and language modeling methods; working with practitioners; building and deploying real systems; evaluating those systems through randomized controlled trials (RCTs) and qualitative interviews; and analyzing regulations and policies. These stages are complementary — a distinctive aspect of my work is in contributing across disciplines: operations, economics, machine learning theory and practice, statistics, and human-computer interaction.
See a recent
"manifesto" on the challenges caused by
heterogeneous participation in participatory systems, which also surveys my work broadly. See a recent
talk video.
Our work has received several awards, including the NSF CAREER, William T. Grant Foundation Scholars Award, INFORMS George Dantzig Dissertation award, ACM SIGecom
Dissertation Award (Honorable Mention), Forbes 30 under 30 for Science, the NSF graduate research fellowship, and paper awards from EC, CSCW, INFORMS, and others. It has been supported by the Sloan Foundation, NSF, NASA, the Cornell Tech Urban Tech Hub, Google, Meta, and
Amazon.
Full bio.
Human-Algorithmic Market Design
Recommenders help us find work, education, and healthcare. I study public interest recommenders, tackling individual-level (e.g., exploration and agency) and societal challenges (e.g., fairness and congestion).
NYC High School matching and recommendation in equilibrium. We work with NYC Public Schools (NYCPS) to address application algorithmic administrative burdens. We first showed (Nature Cities'26) that disadvantaged students disproportionately do not apply to high-performing programs they would have been admitted to had they applied; the gap was largest among the most competitive applicants. In collaboration with NYCPS, we then developed personalized recommendations to reduce disparities, highlighting nearby, high-performing programs where a student had high predicted offer likelihood (ACM EC'26, Best Paper with a Student Lead Author). The central challenge was recommendation-induced congestion: naive recommendations cause sharp, disparate decreases in acceptance rates. We developed and theoretically analyzed a congestion-aware bilevel optimize-and-simulate approach to allocate recommendations and improve match outcomes in equilibrium.
We deployed our approach in the Fall 2025 cycle in an RCT! See the Cornell Chronicle story, and talk video.
Social media feed design. Leveraging Bluesky's open infrastructure, we launched Paper Skygest, a personalized feed for posts about research, with about 1,100 daily active users and over 3 million uses since March 2025. We have a paper describing the feed. Our other projects in the space — a Skytrails research demo and essay — demonstrate a social media experience that allows users to explore their interests.
Platform to help place discharged hospital patients into long-term-care facilities. In Hawaiʻi, a PhD advisee built and deployed a platform to help place discharged hospital patients into one of more than 1,000 long-term-care facilities, many of which are run by single individuals out of their homes. The platform texts homes to ask for updated capacity and preference information, and then provides this information to about 10 hospital social workers; it has helped place hundreds of patients (CSCW'25, Best Paper Award). We also experimentally studied how preferences and incentives shape placements (CSCW'25, Impact Recognition), and then deployed and evaluated active information acquisition with language model support.
Theory. We theoretically model algorithmic monoculture and the wisdom of crowds in matching markets (NeurIPS'24; ACM EC'24) and the downstream strategic, diversity, and fairness implications of ranking algorithms in recommender ecosystems (NeurIPS'24; The Web Conference'24).
Foundations and Systemic Effects of Modern AI
AI helps researchers understand these systems. From my PhD work on text embeddings to modern LLM-based interpretability, I develop language modeling methods to use text as data.
Methods and applications of language models. During my PhD, I developed methods to use word embeddings to study historical societal stereotypes (PNAS'18). LLM-based word embeddings have since dramatically improved, but still face an interpretability challenge: individual vector dimensions are not inherently interpretable, complicating their use. More recently, we used a modern LLM interpretability method — sparse autoencoders —
to tackle this issue (ICML'25). Our method has had considerable follow-up use and contributed to renewed interest in SAEs (ICML'26). We have since begun applying these techniques to social-science questions and studying their theoretical foundations through connections to compressed sensing (COLT'26).
LLM homogeneity and its market implications.
Algorithmic homogeneity is a concern in the LLM age: even as AI systems improve individual decisions, they may homogenize outputs across users,
increasing systemic risk and reducing intellectual diversity. We showed that LLMs make correlated errors, with implications for LLM-as-judge and hiring markets (ICML'25). On the other hand, homogeneity may also be leveraged to detect LLM text in a semi-supervised manner, using test-time adaptation for robustness to distribution shift and strategic behavior (NeurIPS'26).
Diversity and verification in AI-assisted mathematics. Finally, I am working on how to best integrate the progress in AI proof generation into scientific processes. I built an AI workflow for autoformalizing research papers in Lean, formalized most of my own theoretical papers, and launched a AppliedModelingLib — use it to autoformalize your applied math research! (Theory papers I submit now are generally autoformalized using this project.)
Operations and Algorithms for Government
I develop computational methods for efficient and equitable government service allocation.
Resident crowdsourcing, with the NYC Department of Parks and Recreation (Talk video). Do some neighborhoods report more than others for the same underlying conditions on the ground, thus receiving better government services? Answering this question requires new methods because we do not directly observe those conditions. First, we used duplicate 311 reports to estimate reporting rates without external ground-truth data, finding that higher-income and more educated neighborhoods report the same types of problems faster (Nature Computational Science'24). We also leveraged spatial correlation (AAAI'24), combined government ratings and crowdsourced reports (AAAI'26), and used vision-language models (Nature Communications'26) to identify incidents and quantify underreporting.
We then connect measurement to operational decisions. We co-design capacity planning and “service-level agreements:” promises about how quickly incidents of different types will be addressed. We use Bayesian optimization to allocate service capacity across boroughs and incident types, and theoretically analyze efficiency-equity tradeoffs and the gains from centralization (ACM EC'24). We have also built and transferred a dashboard comparing reporting and response patterns and an integer program for NYCDPR's nine-year tree-planting cycle (when each neighborhood is planted), balancing geographic constraints with heat-vulnerability-based need.
Other applications. With the New York Public Library, we analyzed how heterogeneous use of digital holds leads the most desired books to flow from low-income to high-income branches (AAAI'24), and then developed an operational intervention to mitigate this inequity by optimizing from which branch to pull each requested book. We have also used operations research tools for policy design: showing how multi-member districts with ranked-choice voting can curtail partisan gerrymandering (Operations Research'26; ACM EC'22), and studying equity in congestion pricing (ACM EC'24). We further show how optimization tools can be used to debias datasets, from demographic predictions in voter files (PNAS Nexus'25) to granular estimates of intra-US migration (Nature Communications'26).
