Explored model complexity impact on pairs trading strategy performance. Sourced diverse data (Refinitiv API, custom web scraper, Bloomberg), forecasted using 280 features with models ranging from simple regressors to LSTMs. Conducted backtesting using Bayesian-adjusted mean reversion strategy, achieving Alpha of 1+.
Implemented text preprocessing, feature extraction, and multiple clustering algorithms (K-Means, Spectral Clustering). Utilized BERT embeddings and explored dimensionality reduction (PCA, t-SNE) to classify political orientation of news headlines.
Created an interactive Streamlit webapp for US real estate market analysis. Features include interactive maps, statistical visualizations, trend identification, and macro development tracking, enabling users to explore property data and monitor market changes.
Developed a Streamlit app using a multimodal framework to valuate properties. Trained models on data from a large-scale web scraping ETL framework, incorporating Zillow and Yelp data for comprehensive property analysis.
Built a Github Actions based web app deployment framework for EC2 AWS Instances. Utilized PM2 and Nginx for robust process management and reverse proxy functionality.
Web scraper for Zillow, extracting property level data. Utilizes concurrent processing with rotating proxies for efficient large-scale data collection.
Selenium-based Yelp scraper extracting restaurant data. Extracts restaurant information, health ratings, and review counts.
LinkedIn scraper for both company and user profiles. Extracts information including company posts, user experiences, and education.
A collection of notebooks on advanced topics in machine learning and deep learning, including bias-variance tradeoff, logistic regression regularization, neural network architectures, and transfer learning.
Explored quantum computing concepts through algorithmic proofs and Qiskit-based simulations, including the Elitzur-Vaidman bomb detector and quantum game strategies.