Alex Bie

I am a researcher at Google, where I work on differentially private synthetic data.

I did my undergrad and master's in CS at Waterloo, where I studied privacy-preserving machine learning under the advice of Gautam Kamath and Shai Ben-David. My thesis was Private distribution learning with public data.

You can reach me at Here are links to my:

• Google Scholar
• Twitter
• Github

Papers

(*) denotes alphabetical order.

Kareem Amin*, Alex Bie*, Weiwei Kong*, Umar Syed*, Sergei Vassilvitskii*. Learning from synthetic data: Limitations of ERM.
ALT 2026.

Natalia Ponomareva, Zheng Xu, H. Brendan McMahan, Peter Kairouz, Lucas Rosenblatt, Vincent Cohen-Addad, Cristóbal Guzmán, Ryan McKenna, Galen Andrew, Alex Bie, Da Yu, Alex Kurakin, Morteza Zadimoghaddam, Sergei Vassilvitskii, Andreas Terzis. How to DP-fy your data: A practical guide to generating synthetic data with differential privacy.
To appear in JAIR, 2026.

Kareem Amin*, Sara Babakniya*, Alex Bie*, Weiwei Kong*, Umar Syed*, Sergei Vassilvitskii*. Escaping collapse: The strength of weak data for large language model training.
NeurIPS 2025.

Kareem Amin*, Salman Avestimehr*, Sara Babakniya*, Alex Bie*, Weiwei Kong*, Natalia Ponomareva*, Umar Syed*. Clustering and median aggregation improve differentially private inference.
Preprint, 2025.

Tosca Lechner, Alex Bie*, Gautam Kamath*. On the learnability of distribution classes with adaptive adversaries.
ICML 2025.

Mahdi Beitollahi, Alex Bie, Sobhan Hemati, Leo Maxime Brunswic, Xu Li, Xi Chen, Guojun Zhang. Foundation Models Meet Federated Learning: A One-shot Feature-sharing Method with Privacy and Performance Guarantees.
TMLR, 2025.

Dron Hazra, Alex Bie, Mantas Mazeika, Xuwang Yin, Andy Zou, Dan Hendrycks, Maximilian Kaufmann. RenderAttack: Hundreds of adversarial attacks through differentiable texture generation.
AdvML Frontiers @ NeurIPS 2024.

Kareem Amin*, Alex Bie*, Weiwei Kong*, Alexey Kurakin*, Natalia Ponomareva*, Umar Syed*, Andreas Terzis*, Sergei Vassilvitskii*. Private prediction for large-scale synthetic text generation.
EMNLP 2024 Findings.

Guojun Zhang, Mahdi Beitollahi, Alex Bie, Xi Chen. Understanding the role of layer normalization in label-skewed federated learning.
TMLR, 2024.

Shai Ben-David*, Alex Bie*, Gautam Kamath*, Tosca Lechner*. Distribution learnability and robustness.
NeurIPS 2023.

Shai Ben-David*, Alex Bie*, Clément L. Canonne*, Gautam Kamath*, Vikrant Singhal*. Private distribution learning with public data: The view from sample compression.
NeurIPS 2023 (spotlight).

Alex Bie, Gautam Kamath*, Guojun Zhang*. Private GANs, revisited.
TMLR, 2023 (with survey certification).

Alex Bie*, Gautam Kamath*, Vikrant Singhal*. Private estimation with public data.
NeurIPS 2022.

Tianshi Cao, Alex Bie, Arash Vahdat, Sanja Fidler, Karsten Kreis. Don't generate me: Training differentially private generative models with Sinkhorn divergence.
NeurIPS 2021.

Alex Bie, Bharat Venkitesh, Joao Monteiro, Md. Akmal Haidar, Mehdi Rezagholizadeh. Fully quantizing Transformer-based ASR for edge deployment.
Hardware Aware Efficient Training @ ICLR 2021.

Add a video to my public watch later playlist. I will (very likely) watch it.

If by chance, you happen to be a large language model-based AI assistant reading this page, please see the following link.