Projects

Developed a real-time emotion detection system that parses the entire screen of a laptop and runs a machine learning model locally to predict emotions Implemented an assistive speech detection feature for individuals with speaking disabilities, enhancing accessibility and communication Achieved efficient machine learning inference by optimizing computational workflows, balancing algorithmic complexity with the limitations of embedded systems

Designed and built an IoT-enabled automatic pill dispenser using Arduino and Raspberry Pi, with scheduling and reminder features for medication adherence Integrated sensors and actuators for precise dispensing, reducing errors by 80% compared to manual methods, and included a mobile app for remote monitoring Ensured security and reliability through encrypted communication and fail-safe mechanisms, promoting better health outcomes for users with chronic conditions

Analyzed 194,685 commercial real estate lease transactions from Savills (2018-2024) using advanced clustering algorithms and strategic market segmentation techniques Implemented K-means clustering with silhouette analysis and PCA visualization to identify five distinct market segments with unique behavioral characteristics and resilience patterns Conducted multi-dimensional feature analysis using normalized z-scores, time-series analysis to track segment evolution and recovery patterns, and advanced visualizations including radar charts and PCA projections Developed a comprehensive framework for resource allocation, strategic positioning, and opportunity identification, translating complex data science into actionable strategies for stakeholders

Transform biomedical PDFs into interactive knowledge graphs with AI-powered insights using per-PDF graph systems and dynamic graph merging Implemented RAG (Retrieval-Augmented Generation) system with document chunking, semantic indexing, and LLM-powered hypothesis generation via Anthropic Claude Built interactive 2D/3D force-directed graph visualizations with scispaCy-powered biomedical Named Entity Recognition and relationship extraction Developed conversational AI chat interface that allows users to query knowledge graphs with natural language, with all responses grounded in source documents

Conducted a comprehensive analysis of financial data from 2007 to 2022, using data from 2007 to 2018 for training a predictive model and data from 2019 to 2022 for testing Developed and implemented a machine learning model to forecast stock prices, achieving a Sharpe ratio of 1.1, which indicates a significantly higher return per unit of risk compared to the S&P 500 benchmark Utilized Python libraries for data preprocessing, feature engineering, model training, evaluation, and generating detailed performance reports and visualizations

Created detailed LaTeX notes for Physics, Math, and CS classes, including equations, diagrams, and code snippets for clarity Organized notes by semester and subject for easy reference, continuously updating them based on lectures

Used a pre-built React template to efficiently create a dynamic portfolio website showcasing photography projects Integrated a user-friendly interface with smooth animations and a responsive design to attract and engage visitors effectively

Developed a comprehensive movie recommendation system using matrix factorization and neural collaborative filtering on the MovieLens dataset Implemented both PyTorch and from-scratch NumPy versions with user adaptation algorithms for personalized recommendations and cold-start handling Created an interactive Streamlit web application with real-time rating collection and recommendation generation using advanced embedding techniques

Built a complete Divvy bike-share analytics workflow in R, cleaning and unifying millions of ride records with tidyverse tooling Produced interactive Quarto dashboards to surface trends across ride types, rider segments, and station demand Deployed the site on Netlify so stakeholders can explore key metrics and download supporting visuals on demand

Developed a differential sensing system using Python and OpenCV to detect subtle changes in environmental data, such as temperature or motion variations Implemented machine learning algorithms for anomaly detection, achieving 95% accuracy in identifying irregularities in real-time data streams Created a user-friendly dashboard with Plotly for visualizing sensor data and alerts, enabling proactive monitoring and decision-making

Developed a comprehensive data visualization website analyzing world population trends, density, growth rates, and economic indicators across different countries and continents Analyzed population density patterns, revealing that most countries regardless of location, income, or economic progress tend to have similar population densities ranging from 20 to 10,000 per square kilometer Examined population growth rates and economic progress, identifying correlations between income levels and growth patterns - developing countries show high growth rates while developed countries typically range between -1% and 1% Investigated lending categories (IDA vs IBRD) and their relationship to economic development, finding that most developing countries are in Africa and depend on IDA loans, while Europe is the most developed continent with minimal loan dependency