Projects

Director Diffusion

An end to end open-source package to train Low Rank Adaptations (LoRAs) of the Flux1.Krea-dev model to fit the style of famous directors. The directors chosen here are Christopher Nolan, Martin Scorsese, Wes Anderson, Denis Villeneuve, and David Fincher, with the code being broadly applicable.

Motivation / Use Case: Helps those interested in film and the arts better understand the core elements of directorial style. In what ways is it recognizable, and what are subliminal signs? Personally, I was very interested in this due to being a fan of their work, and wanted to understand how to best develop models around them.

Tech / Keywords: Python, Transformers, Torch, CUDA, Modal, Diffusers, XFormers, PeFT, Gradio

Repository: 🔗 GitHub

CourseOdyssey

A personalized platform that allows students to provide their intended degrees (majors, minors, etc.) and planned graduation date, and outputs the set of top-k feasible course plans satisfying those requirements, sorted in descending order on the factor/metric of interest.

Motivation / Use Case: Helps students efficiently plan their academic journey by generating optimized course schedules that satisfy degree requirements while considering their preferred graduation timeline and priorities.

Tech / Keywords: Python, TypeScript, JavaScript, React, Flask, course planning optimization, academic scheduling

Repository: Private due to restrictions with collaborators

PennBook

A miniature Facebook application with posting, news, and chat features, on Node.js and using Amazon EC2 and DynamoDB on the backend. Built in Java and JavaScript.

Motivation / Use Case: Demonstrates social media platform development with real-time features including posting, chat functionality, and personalized news recommendations based on user interactions.

Tech / Keywords: Node.js, Java, JavaScript, Amazon EC2, DynamoDB, Apache Spark, social media, real-time chat, news recommendation system

Repository: Private due to restrictions with collaborators

Surveyor

Surveyor generates data driven surveys for online participants. It uses a Python backend to process surveys and a TypeScript frontend to display them.

Motivation / Use Case: Streamlines the creation and deployment of research surveys by automatically processing survey data and providing an intuitive interface for participants.

Tech / Keywords: Python, TypeScript, CSS, Pug, Pandas, Numpy, NLTK, survey infrastructure, data processing

Repository: 🔗 GitHub