Vinh Le MIT BS '22 & MEng '23

like what you see? repo here!

Vinh Le

Download CV
Access Contact Info

Summary

I am a dedicated and self-motivated software and data engineer with a passion for learning and applying new technologies. I have a lot of experience in creating and innovating systems that drive efficiency and solve complex problems. Being a clear communicator and great team player, I thrive in dynamic environments that allow me to leverage my skills to deliver robust solutions. My track record of successful projects, coupled with my commitment to staying current with industry trends, positions me to align well with any company’s organizational needs to further its success.

Education

  • Massachusetts Institute of Technology(MIT), Cambridge MA,                                                 June 2023

    • Master of Engineering, Computation and Cognition, GPA: 4.2
    • Thesis title: Neuron Image Segmentation via Colorization
      • Advised by: Professor Ernest Fraenkel

  • Massachusetts Institute of Technology(MIT), Cambridge MA ,                                                 May 2023

Related Experience

Analyst                                                                                                       December 2023 - Present
Toyon Research Corporation, Goleta, CA

  • Researched and constructed sparse-tensor based models for classification in dynamic 3-D environments in near-real time.

  • Generated a novel dataset for high-resolution synthetic-aperture radar images using various techniques involving web-scraping, georegistration, and data augmentation.

  • Developed an oriented bounding box detector using Rotated Faster R-CNN for automatic target recognition in synthetic-aperture radar images.

  • Designed and managed projects for multiple interns.
    • Optical character recognition for identifying objects of interest with PaddleOCR.
    • Cylinder detecting in 3-D using surface normals to generate ground-truths.
    • Build of 3-D data collector which made use of a $14k Ouster Sensor and two wide-range cameras.
  • Technologies Utilized:
    • Python, PyTorch, OpenCV, Open3D, scikit-learn, scikit-image, NumPy, scipy, pandas, Docker, Cloud Computing, Linux, SSH, matplotlib, tensorboard, torchmetrics, Jupyter

Research Assistant                                                                                       January 2022 - May 2023
Fraenkel Lab, MIT

  • Constructed and trained a UNet(a type of convolutional neural network) which exceeded state-of-the-art performance on high resolution neuron image segmentation(>98% accuracy).

  • Managed data pipelines efficiently for thousands of high-resolution images, and performed multi-GPU parallel training for rapid prototyping.

  • Developed novel preprocessing techniques and custom model performance metrics to train the model efficiently and accurately present results .

  • Converted a EfficientNet model from Keras to PyTorch, ensuring weights and performance are the same.
  • Technologies Utilized:
    • Python, PyTorch, OpenCV, scikit-learn, scikit-image, NumPy, scipy, pandas, IBM Power9 cluster, Cloud Computing, Linux, SSH, matplotlib, tensorboard, Weights&Biases(wandb), torchmetrics, Jupyter

ML Researcher                                                                                        June 2021 - September 2021
Computational Psycholinguistics Lab, MIT

  • Created reinforcement learning models to represent typing tendencies based on a traditional QWERTY keyboard.
  • Studied recurrent neural networks(RNNs) to construct language models to have human-like biases.
  • Performed data analysis on over 20000 keystroke data points from TypeRacer.
  • Developed probabilistic maps of keyboards to estimate error of typing given a phrase.
    • Technologies Utilized:
      • Python, OpenAI Gym, Keras, scipy, pandas, NumPy, matplotlib, Jupyter

ML Researcher                                                                                             January 2021 - May 2021
Media Lab, MIT

  • Built and benchmarked multiple deep learning models to decide on an optimal inference model for real-time emotion classification.
  • Analyzed audio and video files across multiple datasets to ensure quality of data. Then worked on preprocessing and generating training data.
  • Technologies Utilized:
    • Python, Keras, Tensorflow, scipy, pandas, NumPy, Jupyter

Software Engineering Intern                                                                             January 2020 - May 2020
Micronotes.ai, Boston, MA

  • Produced a Long Short Term Memory (LSTM) network for text-to-text summarization.
  • Applied the model to automated banking interviews to increase efficiency for communication between customers and banks.
  • Attended daily stand-up meetings and followed Agile development practices.
  • Technologies Utilized:
    • Python, Tensorflow, Keras, scipy, NumPy

Computer Science Intern                                                                           October 2019 - December 2019
Research Laboratory of Electronics, MIT

  • Developed data structures to neatly hold information from TextGrid files allowing for efficient cleaning of data.
  • Evaluate and update speech databases that have been labelled for their acoustic cues.
  • Technologies Utilized:
    • Python, scipy, pandas, NumPy, TextGrid, SQL

Events/Publications

Neuron Image Segmentation via Colorization (not available being processed by MIT)                               May 2023

  • My master’s thesis focusing on reframing the problem of neuron segmentation as a colorization problem. This is done by dividing neurons into different “colors”(channels), and having the model distinguish different cells or cell types that way.
  • By utilizing various image processing techniques with torchvision, scipy and OpenCV, I performed transformations on thousands of high resolution images. The model was an Encoder-Decoder structure consisting of ResNet encoder and UNet Decoder. The model was built with PyTorch.

AT dysfunction underlies cognitive deficits in a subset of neuropsychiatric disease models                   August 2021

  • Focused on designing and overseeing experiments collecting mice data.
  • Handled and cared for mice.

Heal-Bot: Quarantine Aide Assistant, Facebook Artificial Intelligence Hackathon                                    June 2020

  • Created an AI chatbot, designed to provide accessible medical information and support. Built using Python, Javascript, html, React, Node, Flask, and various REST APIs.
  • Worked on backend development handling all interactions that occur between user and chatbot. Developed classes to respond in the conversation and gather real time data making use of multiple REST APIs (google-maps, google-geocoding, wit.ai).
  • Trained the wit.ai bot to learn intentions from phrases, which is used for conversation handling.

Core Qualifications

  • Software Development
  • Data Management & Analytics
  • Machine Learning Model Development
  • Testing and Debugging / Rapid Prototyping
  • Teamwork / Clear Communication
  • Version Control and Project Management