Alvin Wan[at]

AI PhD student UC Berkeley

Research Collaborator Facebook Oculus

Google Scholar Github CV

My research produces compact1 and interpretable2 computer vision models for applications like self-driving cars and virtual reality.
1. Efficient Machine Learning involves manually-built neural networks, today. By contrast, our work advances neural architecture search to automatically construct more efficient neural networks, yielding new insights.
2. Explainable AI focuses on saliency, or heatmaps of pixel importance. By contrast, we build interpretable models (with sequential decisions, like decision trees) and extend XAI beyond image classification. (blog)

I’m advised by Joseph E. Gonzalez and participate in RISELab, BAIR, and Berkeley Deep Drive. I received my B.S. ('18) in EECS from UC Berkeley.


  • NSF Fellowship for Machine Learning (2018)
  • UC Berkeley Undergraduate Research Fellowship (2017)


  • Machine Learning: Head TA (Fall '19, Spring '18, Fall '17); TA (Spring '17)
  • Discrete Mathematics and Probability Theory: Head TA (Spring '16, Fall '16, Spring '17)
  • Introductory Computer Science: TA (Fall '15)


  • Facebook Workshop on Neural Architecture Search (April '20)
  • Amazon Graduate Research Symposium (March '19)