Aumit Leon

Software Engineer @ Indigo Ag


Panther Parking

Middlebury College students, faculty, and visitors are often at a loss when it comes to figuring out where they can and can't park. This is due in large part to the confusing collection of parking permits and rules that make it difficult to understand where there is parking available, and if a particular permit is allowed to park in that area. Panther Parking is an interactive informational application that aims to provide insight into where students, faculty, and visitors are allowed to park on campus.

app / repo

Markdown to HTML Converter

A simple to use python module that can convert markdown files into prettified HTML. This project is configured with CircleCI for continuous deployment and uses an opinionated code formatter to enforce consistency between contributors. The module has over 20,000 downloads as of 12/2019.

docs / repo

Accelerating Deep Neural Networks

As part of my undergraduate thesis work at Middlebury College, I explored various parallelization methods aimed at accelerating the training time of Deep Neural Networks. In particular, this work compares the performance of general data parallelism and model parallelism, an expert designed method, as well a generalizable framework on the AlexNet CNN architecture.

report / repo / slides / poster

ArchMLP

Architect Machine Learning Platform (ArchMLP) is an open source platform meant to facilitate the development and deployment of machine learning models through containerization. The platform aims to support data preprocessing, model development, hyperparameter tuning, and model serving.

report / repo

Observing the Evolution of Music with the Million Songs Dataset

A machine learning approach to the Million Songs Dataset in which we utilize Deep Neural Networks and K-Nearest-Neighbor clustering for year prediction, and K-Means clustering to to extract a more refined understanding of the evolution of music features over time.

report / repo

Email Spam Classification

A highly accurate Support Vector Machine model that could classify emails from the preprocessed Enron Spam dataset as spam or ham (non-spam). Achieved 99.6% accuracy, best in class.

report

Increasing EXPEDIA's Bottom Line: Leveraging Packages to Increase Commission

A DataFest 2017 project in which we utilized the Expedia dataset to produce actionable items that would allow Expedia to optimize revenue related to the sale of package deals.

report / repo

Tutoring Forum

A tutoring forum built with the MERN stack. This site allows tutors and teachers to answer important course related questions while reducing the redundancy of questions asked.

repo

see more...


⚒ Built by Aumit Leon
© 2021