Cat P. Le

Machine Learning Scientist

Download Resume Download CV Google Scholar LinkedIn

About Me

Cat P. Le received a B.S. degree Summa Cum Laude in Electrical and Computer Engineering from Rutgers University in 2016, an M.S. degree in Electrical Engineering from California Institute of Technology (Caltech) in 2017, and a Ph.D. degree in Machine Learning from Duke University in 2023. At Duke University, he worked on Task Affinity and Its Applications in Machine Learning under the guidance of Dr. Vahid Tarokh. He currently holds a Postdoctoral Research Associate position at Duke University.

His research interest includes Machine Learning, Computer Vision, and Natural Language Processing, focusing on Transfer Learning, Few-Shot Learning, Continual Learning, Multi-Task Learning, and Neural Architecture Search. His honors include the Matthew Leydt Society, Outstanding Engineering Scholar, John B. Smith Award, Nikola Tesla Scholar, and E. M. Toomey Scholarship. He is also a member of the Tau Beta Pi, Eta Kappa Nu, and Sigma Alpha Pi honor societies.


Duke University

Postdoctoral Research Associate

  • Developing a time-series prediction sytem using TACTis and LLM models.

Duke University

Graduate Research Fellow

  • Developing a novel task affinity based on Fisher Information matrices and maximum bipartite matching algorithm.
  • Task affinity indicates the complexity of transferring the knowledge of one task to another. It is non-commutative and invariant to label permutation.
  • Designing ML frameworks to apply task affinity in Neural Architecture Search, Transfer Learning, Few-Shot Learning, Continual Learning, Multi-Task Learning, Causal Inference, and Image Generative Models.


Research Scientist Intern

  • Analyzing the open-domain dialogs via sentiment analysis, response relevance, response specificity, and text classification.
  • Designing the feature extraction frameworks to utilize the relevant features from the open-domain dialogs for customer rating prediction.
  • Developing a novel open-domain dialog evaluation system based on BERT, LSTM, and causal inference analysis. The model is trained to predict the ratings from the customers and experts.
  • Applying causal inference in the open-domain dialog evaluation system helps improve the flexibility and prediction performance of the model.

Motorola Solutions

Software Engineer

  • Developing the Camera Shutter Synchronization System with LED Strobing for the license plate recognition cameras.
  • Optimizing the Optical Character Recognition (OCR) Algorithm of the license plate recognition cameras.
  • Improving the energy consumption and the detection performance of the license plate and facial recognition cameras.
  • Developing the firmware for the license plate and facial recognition cameras.

California Institute of Technology (Caltech)

Graduate Research Fellow

  • Designing the framework to understand the American sign language using the JPL Sleeve, which receives the signals from 20 hand muscles and maps them into the alphabet.
  • Instructing the lab sections of the Solid-State Electronics for Integrated Circuits course, including fabricating LED, Schottky Diode, MOS Capacitor, P-N Diode, Field Effect Transistor, Bipolar Junction Transistor, Photovoltaic Cell, Microfluidics, and Laser Diode.

Rutgers University

Research Assistant

  • Developing the Radio Access Network under the REU Funding Program of the National Science Foundation (NSF).
  • Designing a simulation with three OpenAirInterface base stations, using USRP B210, which is capable of allocating resources according to users’ demands.
  • Evaluating the data transmission rate between base station (eNB) and users (UEs).
  • Instructing the lab sections of the Digital Signal Processing course.

Wireless Information Network Laboratory (WINLAB)

Summer Research Intern

  • Developing the framework for Wi-Fi, LTE, and LTE in the Unlicensed Spectrum (LTE-U).
  • Design a simulation of the LTE’s base station (eNB) and users (UE) using USRP B210 with the OMF and OpenAirInterface repositories.
  • Evaluating the data transmission rate using the Spectrum, Waterfall, and Constellation plots with varying bandwidths.


Duke University

May 2023

Doctor of Philosophy in Electrical and Computer Engineering

California Institute of Technology (Caltech)

June 2017

Master of Science in Electrical Engineering

Rutgers University

May 2016

Bachelor of Science in Electrical and Computer Engineering

Recent Publications

Improving Open-Domain Dialog Evaluation with a Causal Inference Model

C. P. Le, L. Dai, M. Johnston, Y. Liu, M. Walker, R. Ghanadan

Diversity in Dialogue Systems, IWDSD 2023

Best Paper Award Runner-Up

View Paper

Transfer Learning for Individual Treatment Effect Estimation

A. Aloui, J. Dong, C. P. Le, and V. Tarokh

UAI 2023

View Paper

Task Affinity with Maximum Bipartite Matching in Few-Shot Learning

C. P. Le, J. Dong, M. Soltani, and V. Tarokh

ICLR 2022

View Paper

Fisher Task Distance and Its Applications in Neural Architecture Search

C. P. Le, M. Soltani, J. Dong, and V. Tarokh

IEEE Access Journal, Volume 10, 2022

View Paper

Task-Aware Neural Architecture Search (TA-NAS Framework)

C. P. Le, M. Soltani, R. Ravier, and V. Tarokh


View Paper
Slides Poster

Improved Automated Machine Learning from Transfer Learning

C. P. Le, M. Soltani, R. Ravier, and V. Tarokh

CoRR 2021

View Paper

Supervised Encoding for Discrete Representation Learning

C. P. Le, Y. Zhou, J. Ding, and V. Tarokh


View Paper


Get in Touch