Sara Babakniya

About

I am a Ph.D. student studying Computer Science at the University of Southern California. I am a member of Information Theory and Machine Learning (vITAL) research lab under the supervision of Prof. Salman Avestimehr. Generally, I am interested in different challenges in ML, such as privacy and efficiency. My experience has primarily been exploring these challenges in Federated Learning and Natural Language Processing.

I finished my B.Sc. in Electrical Engineering at the Sharif University of Technology in 2019, during which I gained some experience in implementing and designing networked systems.

Experience

• Student Researcher (Aug 2024 - Present)
  Google Research, New York, NY
  Mentors: Kareem Amin, Umar Syed

• SWE Intern (Feb 2024 - May 2024)
  Google, Sunnyvale, CA
  Mentors: Jeremy Fisher, Sam Aldrin

• Graduate Research Assistant (Aug 2019 - Present)
  University of Southern California, Los Angeles, CA

Publications

• Escaping Collapse: The Strength of Weak Data for Large Language Model Training
  Kareem Amin, Sara Babakniya, Alex Bie, Weiwei Kong, Umar Syed, Sergei Vassilvitskii
  [Authors are ordered alphabetically]
  SSI-FM, ICLR 2025
  [Paper]

• A Data-Free Approach to Mitigate Catastrophic Forgetting in Federated Class Incremental Learning for Vision Tasks
  Sara Babakniya, Zalan Fabian, Chaoyang He, Mahdi Soltanolkotab and Salman Avestimehr
  NeurIPS 2023
  [Paper], [Code]

• Revisiting Sparsity Hunting in Federated Learning: Why does Sparsity Consensus Matter?
  Sara Babakniya*, Souvik Kundu*, Saurav Prakash, Yue Niu and Salman Avestimehr
  Transactions on Machine Learning Research 2023
  [Paper], [Code]

• SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models
  Sara Babakniya*, Ahmed Roushdy Elkordy*, Yahya H Ezzeldin, Qingfeng Liu, Kee-Bong Song, Mostafa El-Khamy, Salman Avestimehr
  FL@FM-NeurIPS 2023
  Best Paper Award
  [Paper]

• AICircuit: A Multi-Level Dataset and Benchmark for AI-Driven Analog Integrated Circuit Design
  Asal Mehradfar, Xuzhe Zhao, Yue Niu, Sara Babakniya, Mahdi Alesheikh, Hamidreza Aghasi, Salman Avestimehr
  ML4S-NeurIPS 2024
  Reproducibility Award
  [Paper]

• Don’t Memorize; Mimic The Past: Federated Class Incremental Learning Without Episodic Memory
  Sara Babakniya, Zalan Fabian, Chaoyang He, Mahdi Soltanolkotab and Salman Avestimehr
  ICML-FL 2023
  [Paper]

• Federated Sparse Training: Lottery Aware Model Compression for Resource Constrained Edge
  Sara Babakniya*, Souvik Kundu*, Saurav Prakash, Yue Niu and Salman Avestimehr
  NeurIPS-FL 2022
  [Paper]

• Supervised Learning for Analog and RF Circuit Design: Benchmarks and Comparative Insights
  Asal Mehradfar, Xuzhe Zhao, Yue Niu, Sara Babakniya, Mahdi Alesheikh, Hamidreza Aghasi, Salman Avestimehr
  Preprint
  [Paper]

• Defending Against Poisoning Backdoor Attacks on Federated Meta-Learning
  Chien-Lun Chen, Sara Babakniya, Marco Paolieri, and Leana Golubchik
  ACM Transactions on Intelligent Systems and Technology, 2022
  [Paper]

• Deep-n-Cheap: An Automated Efficient and Extensible Search Framework for Cost-Effective Deep Learning
  Sourya Dey, Sara Babakniya, Saikrishna C. Kanala, Marco Paolieri, Leana Golubchik, Peter A. Beerel, and Keith M. Chugg
  Springer Nature Computer Science, 2021
  [Paper], [Code]

Honors and Awards

• WiSE Travel Grant, 2024
• Best Paper Award, FL@FM-NeurIPS Workshop, 2023
• Outstanding Poster Presentation, USC MHI Research Festival, 2023
• Best Poster Presentation, USC-Meta Center Workshop, 2022
• Grace Hopper Celebration Travel Grant, USC 2022
• Grad Cohort Travel Grant, CRA-W 2022
• WiSE Qualcomm Top-Off Fellowship 2021
• Grace Hopper Celebration Scholarship 2021

Contact

Email: babakniy[at]usc[dot]edu

University of Southern California
Department of Computer Science
Los Angeles, CA 90089-0781

                                                      last update 2/21/2025