English | Español | Deutsch | 中文

Alex Reyes Aranda

Undergraduate, University of Southern California
Los Angeles, CA
reyesara@usc.edu

Recent News

  • September 2026: I transitioned to a year-round intern with the Secure Algorithms team at Sandia Labs working on Trustworthy AI!
  • May – August 2026: I joined Sandia National Laboratories as an intern in the National Security Programs and worked with the Computer Science Research Institute supporting AI/ML projects part of the Genesis Mission.

I am a first-generation undergraduate computer science student in the CS Theory Group, mentored by the very thoughtful Vatsal Sharan and members of his group. Before USC, I had a great time working with Clare Baek, Dana Saito-Stehberger, and Mark Warschauer in the Digital Learning Lab at the University of California, Irvine.

My interests span optimization, robustness, and learning dynamics for trustworthy machine learning. I research the behavior of machine learning algorithms when they are in adversarial, noise-prone, and resource-limited environments. My work is motivated by the goal of using tools in computer science for developing solutions to societal problems. See brief descriptions of my past projects here.

I actively participate in student-led organizations at USC as the President of SACNAS, and once Programming Chair for Color Stack as well as a Committee Director for SHPE.

Papers

Workshop

On How Muon Reshapes Skill Learning Dynamics
Muon enables more balanced, parallel skill acquisition than SGD and Adam under task-level imbalance, measured by a new Skill Acquisition Lag (SAL) metric.
Alex Reyes Aranda*, Vincent-Daniel Yun, Vatsal Sharan, Bhavya Vasudeva

  • 4th HiLD, International Conference on Machine Learning (ICML), 2026
  • Invited talk at the Annual Machine Learning & Deep Learning Workshop, August 2026 at Sandia National Laboratories

Under Review

Neuromorphic Compiler Optimization
Spectral graph theoretic and load balancing partitioning optimize neuromorphic compiler.
Yang Ho, Felix Wang, Sarah Luca, Alex Reyes Aranda

  • Appearing soon on OSTI.gov

Anticipating Design Tensions Through Early Collective Sensemaking in Research-Practice Partnerships
Early sensemaking as anticipatory design work for navigating equity tensions in RPP curriculum co-design.
Clare Baek, Victoria Nguyen, Symone Gyles, Rossella Santagata, Brian Minh Phan, Alex Reyes Aranda, Mark Warschauer

  • Submitted to the Journal of Learning Sciences, 2026

Projects

Highlighted

Agentic Models & AI Infrastructure
Supporting maintenance on an internal agentic model, and building a locally deployed agentic runtime of my own.
Alex Reyes Aranda*, Chris Siefert, Angel Beltre

    Summer 2026, More

    Phenomenological Deep Learning
    Whitenoise data and spurious patches on CIFAR-10 probe when models heavily rely on simpler features.
    Alex Reyes Aranda*, Siddartha Devic, Julian Asilis, Vatsal Sharan, Shaddin Dughmi, Shang-Hua Teng

      Summer 2025, More

      All past projects can be found here.

Mentorship

Schedule-Stacked RL for Robust Machine Learning
Rowell Pangilinan, Los Angeles Valley College (LAVC)

    Summer 2026, NSF-PURE STEM Award #2413050 LAVC MESA Program

Fellowships

I am gratefully supported by

  • USC Norman Topping
  • Hispanic Scholarship Fund
  • Berkeley Community Scholars

Formerly supported by

  • USC Center for Undergraduate Research in Viterbi Engineering
  • Daben Weiqing Liu Research Cohort
  • Great Minds in STEM, Edison International Scholar
  • National Science Foundation Research Experiences for Undergraduates

Extra

My younger brother is an aspiring physicist: You should check Reese out!