Tijana Zrnic
Stanford University
tijana.zrnic@stanford.edu
Bio
Tijana Zrnic is a Ram and Vijay Shriram Data Science Postdoctoral Fellow at Stanford University, where she is hosted by Emmanuel Candes in the Department of Statistics. Her research establishes foundations to ensure data-driven technologies have a positive impact. Tijana earned her PhD in Electrical Engineering and Computer Sciences from UC Berkeley in 2023, where she was advised by Moritz Hardt and Michael Jordan. Her doctoral research explored prediction and statistical inference in feedback loops, including topics such as performative prediction, prediction-powered inference, and mitigating selection bias. Before her PhD, Tijana completed a BEng in Electrical and Computer Engineering at the University of Novi Sad in Serbia.
Areas of Research
- Machine Learning
Foundations for Reliable and Ethical Data-Driven Decisions
The rapid adoption of data-driven technologies has far outpaced our foundational understanding of them, leaving critical questions unanswered, such as the robustness of these tools under changing conditions and how their performance scales with data and computation. One key set of foundational challenges lies in recognizing that these technologies are not isolated—they are deeply embedded in broader sociotechnical systems, driving high-stakes decisions in policy, science, business, and beyond. For example, black-box AI influences hiring and lending decisions, while large-scale climate models guide climate policies. Without strong foundational principles, such decisions risk being untrustworthy and harmful. Crucially, the solution is not to abandon these technological advances due to gaps in our foundational understanding. Instead, we must forge first principles that meet the demands of current practice while ensuring safe, reliable, and ethical use of these tools in the future. My research establishes foundations to ensure data-driven technologies have a positive impact. I target three major challenges towards this general goal: (1) drawing reliable conclusions from imperfect AI outputs; (2) understanding the interplay between AI and society; (3) building modern foundations of reproducibility. In this pursuit, I draw on interdisciplinary ideas from statistics, computer science, economics, and social sciences. My work is both descriptive and prescriptive, with equal emphasis on both. On the descriptive side, I aim to establish definitions that clarify key challenges, opportunities, and solution concepts in modern data science and AI. For example, I explore what constitutes desirable equilibria between the objectives of AI-powered firms and societal values. On the prescriptive side, I develop reliable methods for modern contexts, such as AI-driven techniques for computing accurate statistical estimates with proper uncertainty quantification. These methods are grounded in strong theoretical foundations, providing much-needed guarantees that help uphold reproducibility and avoid dubious decisions.