Pe'er Lab

Research Assistant (Jan. 2019 - Present)
Advisor: Prof. Itsik Pe'er, Prof. Iddo Drori

  • Developing novel machine learning approaches for high-quality prediction of protein folding / tertiary structure
  • Implemented embedding-based RNN encoder-decoder networks for alpha carbon contact, 3D coordinates, and torsion angle prediction
  • Implemented self-collision checking functionality for protein unfolding dataset generation
  • Published: I. Drori, D. Thaker, A. Srivatsa, D. Jeong, Y. Wang, L. Nan, F. Wu, D. Leggas, J. Lei, W. Lu, W. Fu, Y. Gao, S. Karri, A. Kannan, A. Moretti, C. Keasar, I. Pe'er. "Protein structure prediction with deep learning representations." NeurIPS Learning Meaningful Representations of Life, 2019.

Blei Lab

Research Assistant (Oct. 2018 - Present)
Advisor: Prof. Kriste Krstovski

  • ArXivLab Project: a research paper recommendation system for
  • Developed pipeline for large-scale processing of arXiv data for equation embedding feature extraction
  • Evaluated LDA, doc2vec topic models to optimize research paper recommendation capabilities

NASA Jet Propulsion Laboratory

Help us generate labels for Martian surface features! :)

Deep Learning Technology Group (393K)

Deep Learning Research Intern (May 2019 - Aug. 2019)
Mentor: Dr. Shahrouz R. Alimo

  • Devised CNNs for spacecraft relative pose estimation with multi-scale atrous pyramid pooling and auto-loss weighting
  • Implemented cross-domain spherical embedding encoder-decoder networks for model-based relative attitude prediction
  • Compressed and integrated trained models into low-power embedded systems for on-board inference demonstration

Machine Learning and Instrument Autonomy Group (398J)

Machine Learning Research Affiliate (Sep. 2018 - April 2019)

  • Implemented a VGG16-based Fully Convolutional Neural Network (FCN) that segments input images for detection of Martian Swiss cheese terrain

Machine Learning Research Intern (June 2018 - Aug. 2018)
Mentor: Dr. Lukas Mandrake

  • COSMIC (Content-based Onboard Summarization to Monitor Infrequent Change) Project
  • Devised method to generate high-quality labels for HiRISE browse images of Martian Swiss cheese terrain
  • Performing pixel-wise random forest classification and optimizing performance via feature engineering (e.g. image gradients, filters, spatial information encoding) and hyperparameter tuning
  • Under Review: G. Doran, S. Lu, M. Liukis, L. Mandrake, U. Rebbapragada, K.L. Wagstaff, J. Young, E. Langert, D. Jeong*, A. Trockman*, P. Horton*, A. Braunegg*. "Content-based Onboard Summarization to Monitor Infrequent Change." IEEE Aerospace Conference, 2020.

Columbia Engineering

COMS 6998: Deep Learning Systems Performance Analysis (Dr. Parijat Dube, Fall 2019)

Teaching Assistant (Sep. 2019 - Present)

  • Setting up compute clusters for students, holding weekly office hours, answering questions online, grading assignments, and evaluating final projects

COMS 4721: Machine Learning for Data Science (Prof. John Paisley, Spring 2019)

Teaching Assistant (Jan. 2019 - May 2019)

  • Held weekly office hours, answering questions online, graded assignments and exams
  • Course Website: Click Me

Korea Electronics Technology Institute (KETI)

Researcher (Jul. 2017 - Aug. 2017)

  • Convergence System R&D Division - Artificial Intelligence Research Center
  • Developed a Flask-based web application that records sound (as PCM/WAV) and converts speech to text via Google Cloud Speech API using HTML, JavaScript, and Python (implemented on an Ubuntu Docker container)
  • Presented a seminar on stance detection models involving MLPs, CNNs, gradient-boosted Decision Trees, etc. implemented using Theano and TensorFlow
  • Developed web crawlers using Python to collect Naver/Daum News article data for building an engine that can distinguish fake news articles from real ones
  • Created a demo that displays two different 3D visualizations of the human heart (one generated by a trained AI engine, the other from actual MRI data) for comparison
  • Created training data for sign-language-to-text software development

JRC Robot Inc.

Software Engineering Intern (May 2016 - July 2016)

  • Worked as an Android application developer and database administrator
  • Set up a NAS server and implemented an Android mobile app for remote access for the client firm