Roberto Halpin Gregorio

I am a MS student in the Computer Science Department at Cornell University, studying and performing research in Machine Learning and Computer Vision. I am proud to be advised by Bharath Hariharan and Madeleine Udell. Prior to my MS, I completed my Bachelor of Science degree in Computer Science from Cornell University, specializing in Machine Learning and Computer Vision.

My research and engineering interests lie broadly in understanding, developing, and applying deep learning methods for computer vision and machine learning. Some research areas that I have been involved in throughout my graduate and undergraduate years are representation/self-supervised learning, data augmentation, deep generative models, missing data, amodal/panoptic image segmentation, 3D object detection and tracking, and parallel + distributed machine learning.

I am also proud of being a teaching assistant at the Computer Science Department at Cornell University for over three years. Teaching seven machine learning + computer vision upper-division and graduate semester-long courses.

Currently, I am looking for machine learning (research) engineer opportunities!

Contact: rgh224 [at] cornell.edu

Interests

  • Machine Learning and Computer Vision
  • Science of Deep Learning
  • Self- and Semi-Supervised Learning
  • Deep Generative Models
  • Data Augmentation
  • Object Detection, Tracking, and Segmentation
  • Distributed Machine Learning

Education

  • MS in Computer Science, Present

    Cornell University

    BS in Computer Science, 2020

    Cornell University

projects

project thumbnail
Missing Indicator Method

Introduces Selective MIM (SMIM), a novel method for dealing with missing data.

project thumbnail
Data Augmentation in SSL with GANs

Explores using GANs to generate fake data as a form of data augmentation in self-supervised models.

project thumbnail
MIM-CLR

Develops a self-supervised framework combining contrastive and masked image modeling methods.

project thumbnail
RDMA-wild

Develops RDMA-wild, a novel asynchronous distributed SGD schema for RDMA networks which outperforms synchronous variants over RDMA and TCP.

project thumbnail
Representation Learning Theory

Extension of a previous representation learning theory work, providing more general and robust bounds.

project thumbnail
Pancreas Tumor Segmentation

Addresses the problem of the small amount of available medical data by using 2D-to-3D transfer learning.

teaching

Computer Vision

Spring 2022
CS 4670

Advanced Machine Learning Systems

Fall 2021
CS 6787

Principles of Large-Scale Machine Learning

Spring 2019, 2020, 2021
CS 4787

Foundations of Artificial Intelligence

Fall 2020
CS 4700

Machine Learning for Intelligent Systems

Fall 2019
CS 4780

other

project thumbnail
Lottery Ticket Hypothesis

Overview and summary of the Lottery Ticket Hypothesis.

project thumbnail
Bayesian Deep Learning

Overview and summary of Bayesian Deep Learning.