Vishal Tien

About

Hi! Thanks for visiting my personal website. Here, you can find an assortment of my past projects, as well as a taste of what I'm up to and interested in today.

A little bit about me: I'm a ML Engineer with 2 years of experience in machine learning spanning both research and industry, specializing in natural language processing and with domain expertise in the pharma / healthcare industry. I graduated from the University of Pennsylvania with a bachelors degree in Bioengineering and a master's degree in Systems Engineering with a focus in data science. I am most passionate about NLP, logical / neuro-symbolic AI, and knowledge graphs. Outside of work, I'm a lover of soccer, music, and traveling and am currently located in New York City.

Past Projects

Below are projects I have built outside of work. To see the projects I have worked on in industry, please see my resume

Table of Contents

An Evaluation of Abstractive and Extractive Deep Learning NLP Text Summarization Techniques using Question Answering

Wrote a research paper on a topic of choice for a graduate-level deep learning course at the University of Pennsylvania. Main contributions included implementing beam search decoder as an improvement to Bahdanau's attention-based neural machine translation model fine-tuned on a CNN news articles dataset, as well as introducing a question-answering evaluation mechanism to evaluate and benchmark performance

Conversational Chatbot using Deep Learning

Trained a custom conversational chatbot using TensorFlow and deployed to a web application using Flask. Can be taught to converse about any subject through a simple json file (for example, can currently be asked to retrieve the current Premier League soccer standings).

Implementing a Neural Network From Scratch

Created and trained a neural network from scratch, including manual computation and caching of values from forward process for use in backpropagation to speed up training, using only pure python and numpy. Compared performance to a PyTorch network

Predicting Inpatient Length of Stay at Hospitals

Utilized > 2 million rows of patient data to uncover clinically relevant insights and predict patient length of stay using only data present at time of patient admittance. Makes progress towards improving hospital management efficiency

Image Reconstruction and Generation with Variational Autoencoder

Implemented a variational autoencoder from scratch in the PyTorch library applied to the problem of reconstructing and generating images from MNIST dataset

Machine Learning for Brain-Machine Interfaces

Final project for graduate-level class BE 521: Brain-Computer Interface, where the task was to predict which fingers of a patient were being raised solely from electroencephalogram (EEG) waveform data, with applications to remote neural prosthetic control (for individuals with severe motor disabilities, for example).

Point-of-care Diagnostic Smartphone App for Schistosomiasis Disease using Computer Vision + Microfluidics

Manufactured a low-cost point-of-care diagnostic device powered by a mobile application as part of my team's award-winning senior design project that aimed to reduce the impact and infection of a parasitic disease present in developing countries that are particularly burdened by limited access to healthcare. The Pioneer Award is given to the team that's product best addresses an unmet need as part of the Rothberg Catalyzer event, which is organized with the intent of inspiring the next generation of healthcare innovators

NYC Restaurant Finder Web Application

Developed full-stack React / Node.js web application with Express.js backend connected to a SQL database providing advanced restaurant search capabilities. Click here for full video demo

Ongoing Projects

Trading Bot

An exploration into real time data streaming, threading for concurrency, object oriented programming, and algorithmic trading (genetic algorithms, machine learning, and more) in a custom trading bot

Music Social Media Application with Spotify API

Creating a social media application centered around music. React for frontend and Java SpringBoot backend

Contact

• GitHub
• LinkedIn

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';