Research Positions

BATS Lab | Machine Learning Research Assistant
Mentor: Stephen Bach
Mar 2019 - present

I am currently working on an honors thesis to provide theoretical and statistical explanations for phenomena in weak supervision and self-supervised learning. I have previously worked on a project that developed a performance-guaranteed algorithm for learning from weak labelers, without any assumptions on the distributions of our weak labelers.

Rubenstein Lab | Computational Chemistry Research Assistant
Mentor: Brenda Rubenstein
Nov 2017 - present

My work with Professor Rubenstein applies machine learning methods to study cheminformatics and characterize chemical space. My project generated a novel dataset of structured spectral representations, consisting of NMR, IR, and MS spectra, through the predictions of deep learning models. With this novel dataset, I implemented hierarchical clustering methods to gain an understanding of our molecular representations, comparing its clustering ability to other popular topological fingerprints.

NASA Jet Propulsion Laboratory | Deep Learning Research Intern (393K)
Mentors: Dr. Ryan Alimo, Brian Kahovec
June 2019 - August 2019

My research at NASA JPL looked to improve the detection of failed data transmissions from the Mars Curiosity Rover through various machine learning and deep learning approaches including SVMs, random forests, and logistic regression. I also used unsupervised learning to discover latent structure in the distributions of data transmission and perform anomaly detection with adversarial autoencoders.

MIT Program for Information Science | Data Science Research Assistant
Mentor: Dr. Micah Altman
Sep 2018 - Mar 2019

MIT Program for Information Science | Data Science Research Intern
May 2018 - Sep 2018

I worked on the CensusUsage project, which was an open source research project aiming to categorize different use cases of the American Community Survey (ACS) data. Using the datasets that I had created, I analyzed the corpuses using unsupervised topic models based on LDA and NMF to identify articles and topics that fall within our pre-determined use cases. I implemented a LSTM as an alternative method to classify articles' use cases in a supervised fashion.