Portfolio

PocketChange

This startup provides a secure, end-to-end platform that simplifies gift card liquidation, trading, and storage, backed by a trading algorithm, API-based financial SaaS integrations, and advanced ML recommendations. Having raised over $50K in funding and achieved top-tier recognition with YC W25 and Techstars NYC final-round interviews, it has also secured LOIs with major brands like McDonald’s, Dunkin, and PDQ. Built with technologies like Dart, Python Django, AWS, and Docker, the solution ensures scalable, secure transactions and seamless gift card management for both users and corporate partners.


LegalEval Large Language Model: Understanding Legal Texts

Developed a custom PyTorch data loader that enhanced LLaMA2 model accuracy by 15% for log probability analysis, showcasing expertise in optimizing deep learning workflows. Leveraged Hugging Face Accelerate and DeepSpeed ZeRO 3 to implement model parallelism, reducing training time by 30% on A100 GPUs. Applied these advancements to legal text classification tasks for SemEval 2023 LegalEval, including Rhetorical Roles Labeling and Court Judgment Prediction, delivering impactful results in natural language processing.


Convolutional Neural Network Bird Classifier

Led the development of a custom bird classification model in PyTorch, achieving 87% accuracy on a dataset of 3,113 bird images. Utilized data augmentation and GAN-based image generation techniques to reduce overfitting by 20%, enhancing model generalization. This project demonstrated proficiency in deep learning and image classification, addressing challenges in dataset variability.


Supreme Court LLM Data Trainer

Conducted advanced statistical analyses on over 5,000 survey entries with 150+ variables to assess public opinion’s influence on Supreme Court decisions, employing chi-squared tests, linear regression, and confidence intervals. Developed data visualizations in R, Stata, and Python to explore SCOTUS docket trends, case alignments, and political beliefs of small businesses. Collaborated with Dr. Neil Malhotra on a research publication and a forthcoming book, “Majority Opinions: How an Out-of-Step Supreme Court Can Affect the Rule of Law,” while fine-tuning AI models leveraging LLMs for predicting Supreme Court decisions.


Matrix Theory Linear Algebra Algorithms

In Matrix Analysis for Scientists and Engineers by Alan J. Laub, algorithms are designed to provide efficient computational techniques for linear algebra operations, such as matrix decompositions, solving linear systems, and eigenvalue computations. Building on these concepts, I developed Python implementations of key linear algebra algorithms and matrix formulas foundational to machine learning models. These include Principal Component Analysis (PCA), Singular Value Decomposition (SVD), linear and logistic regression, least squares optimization, and k-means clustering, enabling efficient data processing and insights into high-dimensional datasets.


Wikiscraper

Developed a novel program in C++ to find link ladders between two Wikipedia pages using an intelligent algorithm to identify optimal paths. This project involved leveraging core C++ features such as iterators, algorithms, and containers to efficiently traverse and analyze graph-like structures. The program not only enhanced problem-solving and algorithmic design skills but also deepened expertise in fundamental C++ concepts.


Soap Store

OnlineSoapShop is a Java web application designed to facilitate the sale of soap for a fundraiser, leveraging databases to store customer information and product lines. Utilized Sprint Boot, SQL, Postrge, Postman, Docker, and many other SWE applications. The project is primarily developed in Java and includes features such as user registration, product catalog, shopping cart, order tracking, and an admin panel for managing products and orders.


Weather App

This project is a full-stack weather application that allows users to retrieve current weather information for any location using a weather API. The app combines a JavaScript-based front-end with a Node.js-powered back-end. NPM is utilized for managing dependencies, and Java is incorporated for additional processing, such as complex calculations or integrations, providing a robust and scalable solution.