Andrej Risteski

Andrej Risteski is an Associate Professor at the Machine Learning Department of Carnegie Mellon University (CMU) since 2025. He previously held the Norbert Wiener Research Fellow position jointly between the Applied Math Department and IDSS at MIT (2017–2019) after completing his PhD in Computer Science at Princeton University (2012–2017) under Sanjeev Arora. His research focuses on the intersection of machine learning, statistics, and theoretical computer science, emphasizing generative models, representation learning, and out-of-distribution generalization with applications to natural language processing and scientific domains.

His recent publications explore edge embeddings in Graph Neural Networks (GNNs), score matching efficiency, and theoretical foundations of diffusion models. Key contributions include analyzing computational bottlene.com/activities/statistical-and-computational-challenges-in-probabilistic-scientific-machine-learning-sciml/">NSF CAREER Award, DOE Computational Science Graduate Fellowship for Stephen Huan, and co-organizing the COLT workshop on Theory of AI for Scientific Computing. He advises PhD students across Machine Learning, Computer Science, and Mathematics, including Bingbin Liu (Kempner Institute Fellow) and Elan Rosenfeld (Google Research Scientist). Teaching includes Probabilistic Graphical Models and Advanced Deep Learning at CMU, plus Applied Mathematics at MIT. His work is supported by NSF, DoD, and OpenAI Superalignment grants.

• Education: PhD in Computer Science (Princeton), BSc in Computer Science (Princeton)
• Current Positions: Associate Professor, CMU Machine Learning Department
• Former Positions: Norbert Wiener Fellow, MIT IDSS & Applied Mathematics

Research Areas:

• Generative Models (GANs, Diffusion Models)
• Representation Learning
• Out-of-Distribution Generalization
• Neural Language Models
• AI for Scientific Applications
• Sampling and Optimization Algorithms

Scientific Awards:

• NSF CAREER Award (2023)
• Google Research Award (2024)
• Amazon Research Award (2022)
• OpenAI Superalignment Grant (2023)

Recent Talks (2023–2025):

• "Architectural Nuances and Benchmark Gaps in Scientific ML" (UC Berkeley, 2025)
• "The Statistical Cost of Score-Based Losses" (Simons Institute Boot Camp, 2024)
• "Neural Networks for PDEs" (ETH Zurich, 2024)
• "Discernible Patterns in Transformers" (Theory of Interpretable AI, 2024)

He leads a research group producing work at the interface of computational complexity and graph learning, with empirical validation on topological bottlenecks and hub node dynamics. Current projects include ICML 2025 paper on edge embeddings in GNNs and COLT 2025 workshop on AI for Scientific Computing co-organized with MIT, Duke, and ETH Zurich collaborators.