Sushant Agarwal

I am a graduate student in the Computer Science department at Northeastern University, where I am advised by Ravi Sundaram and Rajmohan Rajaraman. Prior to this, I graduated with a Master's degree in Computer Science from the University of Waterloo, where I was advised by Shai Ben-David. Previously, I did my Bachelor's in Mathematics and Theoretical Computer Science from Chennai Mathematical Institute.

I am broadly interested in theoretical machine learning, and have recently been focusing on ethical issues that arise in machine learning models, such as fairness, interpretability, privacy, and robustness. I am motivated by problems that are theoretically challenging, while also being practically relevant.

Email  /  Google Scholar  /  Twitter

profile photo
Publications
An Efficient Geometric Characterization of Robust Fair Learning
Sushant Agarwal, Amit Deshpande, Rajmohan Rajaraman, Ravi Sundaram
In Submission

Aggregating Data for Optimal and Private Learning
Sushant Agarwal, Yukti Makhija, Rishi Saket, Aravind Raghuveer
UAI 2025 (Oral Presentation) [pdf]
  • Presented at the Google Research Conference on Ads Privacy 2025
  • Presented at the NeurIPS 2024 workshop on Optimization for Machine Learning
Optimal Fair Learning Robust to Adversarial Distribution Shift
Sushant Agarwal, Amit Deshpande, Rajmohan Rajaraman, Ravi Sundaram
ICML 2025 [pdf]
  • Presented at FORC 2025 (poster)
Private Mean Estimation with Person-Level Differential Privacy
Sushant Agarwal, Gautam Kamath, Mahbod Majid, Argyris Mouzakis, Rose Silver, Jon Ullman
SODA 2024 [pdf]
  • Presented at TPDP 2025
On the Power of Randomization in Fair Classification and Representation
Sushant Agarwal, Amit Deshpande
FAccT 2022 [pdf]

Towards the Unification and Robustness of Post-hoc Explanations
Sushant Agarwal, Shahin Jabbari, C. Agarwal*, S. Upadhyay*, Hima Lakkaraju, Steven Wu (CO)
ICML 2021 (Spotlight Presentation) [pdf]
  • Presented at FORC 2022 (non-archival track) [pdf]
Open Problem: Are all VC-classes CPAC learnable?
Sushant Agarwal, Nivasini A., Shai Ben-David, Tosca Lechner, Ruth Urner
Open Problems @ COLT 2021 [pdf]

On Learnability with Computable Learners
Sushant Agarwal, Nivasini A., Shai Ben-David, Tosca Lechner, Ruth Urner
ALT 2020 [pdf]

On Trade-offs between Fairness, Interpretability, and Privacy in Classification
Sushant Agarwal
Master's Thesis [pdf]
  • Chapter 3 presented at the AAAI 2021 workshop on Explainable Agency in AI [pdf]
  • Chapter 3 presented at the IJCAI 2021 workshop on AI for Social Good [pdf]
  • Chapter 4 presented at the IJCAI 2021 workshop on AI for Social Good [pdf]
Impossibility Results for Fair Representations
Tosca Lechner, Shai Ben-David, Sushant Agarwal, Nivasini A. (CO)
[arxiv]


Authorships are in alphabetical order, unless indicated. CO denotes contributional order. * indicates equal contribution.
Thanks to Jon Barron for the template.