Shivvrat Arya is the Principal Investigator and Director of the Algorithms and Architectures for Reasoning and Intelligent Automation (ARIA) Lab. He earned his Ph.D. in Computer Science from The University of Texas at Dallas, where he was advised by Vibhav Gogate and Yu Xiang.

His research focuses on interpretable, trustworthy, and efficient artificial intelligence, with an emphasis on neurosymbolic reasoning, probabilistic inference, and learning-based approaches to combinatorial and constrained optimization. He develops neural and hybrid reasoning systems that integrate structured knowledge with data-driven learning to enable reliable decision-making in complex environments. His work has advanced neural inference, structured reasoning, and optimization methods, with applications in computer vision, video understanding, multimodal reasoning, and human-centered AI. His contributions have been recognized with Best Paper, Spotlight, and Oral Presentation Awards at major AI venues, including NeurIPS and AAAI.

He is recruiting motivated PhD, MS, and undergraduate students. Hiring details are available at https://aria-research-lab.github.io/hiring/.

Ph.D. ; THE UNIVERSITY OF TEXAS AT DALLAS, RICHARDSON, TEXAS ; Computer Science ; 2025

M.S. ; THE UNIVERSITY OF TEXAS AT DALLAS, RICHARDSON, TEXAS ; Computer Science ; 2021

B.Tech. ; INDIAN INSTITUTE OF INFORMATION TECHNOLOGY VADODARA, INDIA ; Computer Science and Engineering ; 2019

https://shivvrat.github.io/

Neurosymbolic AI & Explainable Systems: Integrating symbolic structure with deep learning for transparent reasoning. We develop hybrid architectures that combine neural pattern recognition with logical inference for interpretable decision-making.

Neural Combinatorial Optimization: Learning-based solvers for combinatorial and constrained problems. Our work explores neural architectures that learn to solve NP-hard optimization problems efficiently.

Deep Reinforcement Learning for Graph Optimization: Graph neural networks combined with reinforcement learning for solving complex graph-based optimization problems, including routing, scheduling, and resource allocation.

Probabilistic Inference: Scalable neural methods for reasoning under uncertainty. We develop efficient algorithms for probabilistic graphical models and neural approaches to approximate inference.

Applications of Neurosymbolic Methods: Computer vision, video understanding, activity recognition, human-computer interaction, and multimodal reasoning. Applying neurosymbolic AI to real-world tasks requiring structured understanding.

"Learning to Condition: A Neural Heuristic for Scalable MPE Inference"
arXiv preprint arXiv:2509.25217, 2025.

"SINE: Scalable MPE Inference for Probabilistic Graphical Models using Advanced Neural Embeddings"
2025.

"RELINK: Edge Activation for Closed Network Influence Maximization via Deep Reinforcement Learning"
ACM, November, 2025.

"A Neural Network Approach for Efficiently Answering Most Probable Explanation Queries in Probabilistic Models"
Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2024.

"CaptainCook4D: A Dataset for Understanding Errors in Procedural Activities"
Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2024.

Chiradeep Roy, Mahsan Nourani, Shivvrat Arya, Mahesh Shanbhag, Tahrima Rahman, Eric D. Ragan, Nicholas Ruozzi, Vibhav Gogate. 2023. "Explainable Activity Recognition in Videos using Deep Learning and Tractable Probabilistic Models." ACM Transactions on Interactive Intelligent Systems , vol. 13 , no. 4 , pp. 1-32.