I am on the 2026–27 academic job market.
I am a U.S. permanent resident and can work in the United States without visa sponsorship.
What has to be true about an AI system before anyone can rely on it?
About
I am a Postdoctoral Research Associate at Stanford University, in the Center for Sustainable Development and Global Competitiveness, working with Prof. Jie Wang and Prof. Michael Lepech. I was previously an Associate Research Scientist at Columbia University, and I received my Ph.D. in Computer Science from the University of Virginia in 2024, advised by Prof. Bradford Campbell. My research builds AI systems that can be trusted with impactful decisions, across four connected areas:
- Trustworthy generation and agents: generation whose output can be verified, from code to clinical text.
- Alignment at large scale: alignment and calibration methods that hold under continual distribution shift.
- Privacy-preserving and efficient learning: federated learning and efficient inference where data cannot leave the device.
- High-stakes deployment: healthcare, the built environment, finance, etc., where errors are consequential.
My work appears at ICML, NeurIPS, ICLR, AAAI, KDD, WWW, SIGIR and ACM MM, in the ACL family, and at systems venues including ICCPS, IPDPS, SenSys and IPSN, and it has been recognized by a Google HE Faculty AI Fellowship, a Best Paper Nomination at the EMNLP Industry Track, orals at ACM MM and CHIL, a Best Paper Candidate at ACM BuildSys, and the World's Top 2% Scientists list. I also treat shared evaluation resources as part of the research, contributing to the FinRL open-source ecosystem and organizing the FinRL Contests.
Research overview
A closer look at each line, with the papers I would point to first:
Trustworthy Generation & Agents
Output that can be checked. Generated code has to compile and run across frameworks, clinical text has to be grounded in anatomy, and both need evaluation infrastructure that decides whether they hold.
Alignment at Large Scale
Alignment and calibration tested in large-scale recommendation platforms, where distributions shift daily and every retraining run has a cost.
Privacy-Preserving & Efficient Learning
Learning where the data and the compute actually are. Data stays on the device or in the hospital, budgets are finite, and the accuracy cost of both has to be part of the design.
AI for High-Stakes Domains & Open Source
The places that do not forgive. Hospitals, buildings and energy systems, and financial markets, together with the open-source ecosystems that carry the work past the paper.