Adya Agrawal

756 W Peachtree St NW, Atlanta GA 30308

I am a first year PhD student in the College of Computing at Georgia Tech, where I am advised by Prof. Vassilis Zikas. My research primarily focuses on privacy-preserving machine learning and applied cryptography. I am currently working on privacy for active learning and recommendation systems.

Prior to starting my PhD, I was a software engineer at JPMorgan Chase and Co. where I was contributing to research projects with the J.P. Morgan AI Research team, particulary the AlgoCRYPT COE under the mentorship of Dr. Antigoni Polychroniadou. I graduated with a Bachelors of Technology (B.Tech) in Computer Science from Vellore Institute of Technology, India in July 2023.

I am broadly interested in enhancing privacy and security in real-world AI systems. I aim to integrate privacy-enhancing technologies (PETs) in distributed systems to protect sensitive data, while maintaining practical utility and accuracy.

Google Scholar / LinkedIn / Twitter / Github / Medium

Selected Publications

Pre-Print

Let’s Ask Gauss: Improved One-Run Privacy Auditing

Adya Agrawal, Yu Wei, Jaspal Singh, Malik Magdon-Ismail, and Vassilis Zikas

2026

Submitted to NeurIPS 2026.

Abs arXiv Code

Privacy auditing provides an important safeguard by estimating the actual information leaked by a model, thus ensuring that theoretical privacy guarantees hold in practice. We study empirical privacy auditing for differentially private (DP) machine learning, focusing on efficient one-run methods for mechanisms such as DP-SGD. Prior one-run approaches threshold training examples or "canaries" into binary membership guesses, which discards useful information. We show that, in the white-box DP-SGD setting, canary-aligned signals naturally form a sequence of random variables whose normalized sum is asymptotically Gaussian. Leveraging this distributional perspective, we develop a DP-auditing framework that leads to tighter privacy lower bounds from a single training run.
Pre-Print

Balanced Additive Randomized Encodings for Shuffle Differential Privacy

Yu Wei, Jaspal Singh, Adya Agrawal, and Vassilis Zikas

2026

Presented at TPDP 2026. Submitted to ACM CCS 2026.

PDF Code
ICML
EncryptedLLM: Privacy-Preserving Large Language Model Inference via GPU-Accelerated Fully Homomorphic Encryption

Leo Castro, Daniel Escudero, Adya Agrawal, Antigoni Polychroniadou, and Manuela Veloso

In Forty-second International Conference on Machine Learning, 2025

Abs Bib PDF Code Website

As large language models (LLMs) become more powerful, the computation required to run these models is increasingly outsourced to a third-party cloud. While this saves clients’ computation, it risks leaking the clients’ LLM queries to the cloud provider. Fully homomorphic encryption (FHE) presents a natural solution to this problem: simply encrypt the query and evaluate the LLM homomorphically on the cloud machine. The result remains encrypted and can only be learned by the client who holds the secret key. In this work, we present a GPU-accelerated implementation of FHE and use this implementation to benchmark an encrypted GPT-2 forward pass, with runtimes over 200x faster than the CPU baseline. We also present novel and extensive experimental analysis of approximations of LLM activation functions to maintain accuracy while achieving this performance.
@inproceedings{castro2025encryptedllm, title = {Encrypted{LLM}: Privacy-Preserving Large Language Model Inference via {GPU}-Accelerated Fully Homomorphic Encryption}, author = {de Castro, Leo and Escudero, Daniel and Agrawal, Adya and Polychroniadou, Antigoni and Veloso, Manuela}, booktitle = {Forty-second International Conference on Machine Learning}, year = {2025}, url = {https://openreview.net/forum?id=PGNff6H1TV}, }