Tech Stack: Gradio, Python Imaging Library (PIL), Javascript

A fast and lightweight image editor natively integrated with SD Web UI, the largest open-source ecosystem for AI image generation. This tool helps speed up AI image generation workflows for all platform users

Tech Stack: Python, Flask, RegEx Reddit API, Praw.

A Reddit bot that converts temperature values and makes some jokes.

(Constantly praised by English-speaking Redditors while getting screamed at by French and German ones)

Use Low-rank Adaptation (LoRA) to finetune a Large Language Model (LLM) on course materials to create a QA chatbot. This AI can answer questions about the study materials to help students study for exams.

Tools: Huggingface transformers, Pytorch

Develop machine learning models for locating rice fields, leverage satellite data using Microsoft Planetary Computer Hub.

Achieve F1 score of 0.95, top 61 on the global leaderboard

Develop a chess engine using the MCTS algorithm

Develop deep learning models (based on pretrained CNN such as DenseNet) to detect and classify crimes from security camera footage. Model achieves AUC score of 0.83

Final project paper for CISC6000 Deep Learning, Fall 2022.

Develops a fully functioning neural network from scratch using only Python and NumPy. Network can contain multiple layers with any number of neurons, process forward pass and backpropagation for training using Stochastic Gradient Descent (SGD) optimizer with Sigmoid activation function. Optimizer SGD is upgraded to incorporate training with momentum, which speeds up the training proceess for faster convergence.

Final project for CISC5800 Machine Learning, Spring 2022.

Use deep convolutional neural networks to classify hand-written digits from the MNIST dataset.

Final project paper for CISC4631 Data Mining, Fall 2021.