CV

Education

PhD, University of California San Diego

  • September 2021 - Current
  • PhD student in Computer Science and Engineering, advised by Prof. Lily Weng

B.Sc, Massachusetts Institute of Technology(MIT)

  • September 2017 - June 2021
  • Bachelor of Science in Computer Science and Engineering and in Philosophy
  • GPA 5.0/5.0, Minor in Mathematics

Work Experience

UC San Diego, Research Assistant

  • September 2021 - Current
  • Developing scalable methods to deeply understand neural networks, or create neural networks that are easier to understand, in order to improve their safety and reliability.
  • Mostly focused on explaining individual neurons in vision and language model, and developing methods to make them more interpretable.

IBM, Research Intern at MIT-IBM Watson AI Labs

  • June 2022 - September 2022
  • Developed Label-free Concept Bottleneck Models, a novel method to create understandable neural networks. The method allows us to turn any pretrained network into a CBM where each neuron in the second to last layer corresponds to a human understandable concept. The method requires no labeled concept data and retains high performance. Published results in ICLR 2023.

MIT Department of EECS, Undergraduate Researcher

  • Sep 2019 - May 2021
  • Developed an effective framework for training reinforcement learning agents that are robust against adversarial attacks on observations, works against larger perturbations and is around 5 times faster than previous approaches without losing natural performance, published at NeurIPS 2021

Vectra AI, Data Science Intern

  • Summer 2019, 2020
  • Created GraphMDN, a combination of graph neural networks and mixture density networks for solving inverse problems on graph structured data. Published results on human pose estimation.
  • Created tools for better comparison and parallel execution of machine learning models in real world cybersecurity settings and used these to improve the performance of a production algorithm by 50%.

Other

Programming Languages: Very experienced with Python and Pytorch, some experience with TensorFlow, Java

Awards: International Physics Olympiad: Silver medal 2016, Bronze medal 2015

Reviewer: NeurIPS 2022-23, ICML 2023, ICLR 2024