Featured Projects
Real-world data science work that drives business decisions
Cultural Pulse Predictor
Building an end-to-end, production-grade ML pipeline that ingests real-time data from Spotify charts, Billboard, Reddit, TikTok trends, and news headlines to predict emerging artists and cultural moments before they break mainstream. Features engineered from audio characteristics and social signal velocity detect organic vs. artificially accelerated momentum.
Impact: Full-stack ownership — pipeline architecture, feature engineering, ML modeling, and dashboard deployment.
Streaming vs. Theatrical Analytics
End-to-end analytics pipeline analyzing 530 films (2019-2024) using TMDb API, SQL, and Tableau. Discovered Action films achieve 2,437% ROI theatrical-only, while Crime films gain 582% from streaming releases.
Impact: Insights could inform $10-50M release windowing strategies for major studios. View Live Dashboard →
Regional Music DNA
Analyzed regional music preferences across major US cities by integrating Spotify Web API, Last.fm API, and US Census data to identify geographic patterns in musical taste.
Impact: Enables hyper-local marketing strategies for streaming platforms and concert promoters.
Enterprise Data Governance @ Sallie Mae
Supported Chief Data Office in implementing enterprise-wide data governance initiatives. Built Python API integrations and enhanced metadata accuracy in Alation.
Impact: Strengthened data accessibility and decision-making across the organization.
GA4 E-Commerce Business Insights
SQL-based analysis of Google Analytics 4 e-commerce data (270K users, $362K revenue) uncovering critical retention crisis: 93.6% customer churn after first purchase. Analyzed cohort behavior, marketing attribution, and revenue drivers to deliver actionable recommendations.
Impact: Identified $17.7K+ revenue recovery opportunity through Month 1 re-engagement campaigns and optimized marketing channel allocation.
Cookie Cats: A/B Test Analysis
Analyzed 90K mobile game players to test gate placement impact on retention (Day 1: 44.8% vs 44.2%, Day 7: 19.0% vs 18.2%). Used bootstrapping and statistical testing in R to determine optimal player experience strategy.
Impact: Provided data-driven recommendation on feature placement affecting user retention. View Full Analysis →
Image Classifier with Transfer Learning
Built image classification model using transfer learning (ResNet18 & VGG13) to classify 102 flower species with 60%+ accuracy. AWS Scholarship project.
Skills: PyTorch, Transfer Learning, GPU Optimization.
Altristotle — Custom LLM for Student Learning
Built a customized LLM that guides students toward solutions through Socratic questioning rather than providing direct answers, designed for underfunded schools. Developed during the MIT RAISE Create-a-Thon (SureStart program) and deployed locally on GPU-supported environments for real-time interaction.
Skills: Python, Ollama, WebUI, LLM fine-tuning.