Kameron Shahabi

About

I am a master's student in computer science at USC. I am part of the Theory Group, advised by the brilliant Prof. Vatsal Sharan.

My interest lies in theoretical machine learning, specifically robustness, fairness, and generalization. I am also fond of cryptography, which I studied during my research internship at Signal research. I am currently thinking about the following questions:

• How can we theoretically understand learning in the presence of spurious features? Are models that rely on these features despite achieving low training error memorizing datapoints? Can we draw connections to multicalibration?
• Can we explore the implicit biases of gradient-based methods from the lens of stability? Under which settings is gradient descent the most stable learning algorithm?
• Can we scale current private information retrieval (PIR) systems to implement secure contact discovery for a commercial user-base (Signal)?

I am currently applying to Ph.D. programs to start in Fall 2024. Please reach out if there is anything you would like to discuss!

Publications

• Mitigating Simplicity Bias in Deep Learning for Improved OOD Generalization and Robustness
  Bhavya Vasudeva, Kameron Shahabi, Vatsal Sharan
  ICML Spurious Correlations, Invariance, and Stability, 2023
  ICML SCIS | pdf

• Toward Accurate Spatiotemporal COVID-19 Risk Scores Using High-Resolution Real-World Mobility Data
  Sirisha Rambhatla, Sepanta Zeighami, Kameron Shahabi, Cyrus Shahabi, Yan Liu
  ACM Transactions on Spatial Algorithms and Systems, 2022
  ACM TSAS | pdf

Projects

• pir-psi-contact-discovery: Prototype implementation of the private set intersection scheme from the paper 'Scaling Mobile Private Contact Discovery to Billions of Users'.

Education

• M.S. in Computer Science, University of Southern California
  Expected May 2024

• B.S. in Computer Science, University of Southern California
  May 2023, Summa Cum Laude