Alex Gittens
Lally Management Building, Office 316
110 8th St
Troy, NY 12180
I am an associate professor of computer science at Rensselaer Polytechnic Institute. My research focuses on designing algorithms with provable performance guarantees, with an emphasis on tensor factorization, trustworthy machine learning, and randomized numerical linear algebra for large-scale machine learning applications.
I have advanced scalable methods for tensor decomposition, developed robust techniques for scalable and fair machine learning, and explored randomized algorithms to optimize computational and communication efficiency in data-intensive and distributed problems. My work has been published in top-tier venues and applied across domains, from scientific computing to real-world machine learning systems.
Before joining RPI, I earned my PhD in applied and computational mathematics from CalTech and held research positions at eBay, the International Computer Science Institute, and UC Berkeley’s AMPLab.
selected publications
- Patching LLMs Like Software: A Lightweight Method for Improving Safety Policies in Large Language ModelsWorkshop paper ICLR 2026 Trustworthy AI, 2025arXiv:2511.08484
- Exploiting Missing Data Remediation Strategies Using Adversarial Missingness AttacksIn Fortieth AAAI Conference on Artificial Intelligence, 2026
- Group Fair Federated Learning via Stochastic Kernel RegularizationTransactions on Machine Learning Research, 2025
- PEEL the Layers and Find Yourself: Revisiting Inference-Time Data Leakage for Residual Neural NetworksIn 2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), 2025
- Aligners: Decoupling LLMs and AlignmentIn Findings of the Association for Computational Linguistics: EMNLP 2024, Nov 2024
- Simple Disentanglement of Style and Content in Visual RepresentationsIn International Conference on Machine Learning, ICML 2023, Nov 2023
- Deception by Omission: Using Adversarial Missingness to Poison Causal Structure LearningIn Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2023, Nov 2023
- Adaptive Sketching for Fast and Convergent Canonical Polyadic DecompositionIn Proceedings of the 37th International Conference on Machine Learning, ICML 2020, Nov 2020
- NoisyCUR: An Algorithm for Two-Cost Budgeted Matrix CompletionIn Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Nov 2020