Data Scientist
Boston, Massachusetts
I build production ML systems that drive real decisions. From forecasting models predicting daily volume across dozens of sites to cloud-deployed scientific computing platforms, I turn data into tools people actually use.
I like building tools that make people's jobs easier, predicting the future so businesses are prepared, and doing research that pushes things forward. Outside of work, I play Europa Universalis IV. Feel free to reach out if you want to talk about data, collaborate on something, conquer the HRE together, or just sign my guestbook.
Currently: Graduated with my M.S. in Data Science from UVA (Aug 8, 2026), co-authoring a molecular docking paper with the Trinh Lab, and preparing for the AWS Certified Generative AI Developer exam. I also write about my work on the blog.
Production ML system predicting daily IT contact volume across multiple corporate sites. GradientBoosting achieves R-squared 0.94+ on high-volume sites. Used by management to shift from reactive to proactive staffing.
End-to-end ML pipeline forecasting daily trading volume for the S&P 100. Compares 4 models per ticker — XGBoost wins with 3-7% MAPE. Auto-retrains daily after market close.
Real-time fraud detection with sub-3ms inference, Redis feature store for per-card velocity tracking, and a live monitoring dashboard. XGBoost classifies 590K transactions with 64% recall at 35% precision.
Reframed IMDb sentiment as a 3-class problem. Evaluated 5 architectures (LSTM, BiLSTM, CNN, GRU, BERT). BERT achieved macro-F1 0.784 with systematic ablation studies.
NLP pipeline analyzing 12 novels from Project Gutenberg. Tokenization, TF-IDF, PCA, LDA topic modeling, and word2vec revealed cross-genre cultural evolution patterns.
Multiple statistical models classifying potential survivors based on shelter conditions. Compares logistic regression, LDA, and tree-based methods.
Interactive dashboard exploring salary disparities across demographics using General Social Survey data. Built with Plotly Dash for real-time filtering and visualization.
Visitors ask questions about my background and get instant answers. RAG pipeline retrieves relevant context from a knowledge base, generates responses via LLaMA 3.1 on Groq.
Automated daily scraper tracks which skills employers actually ask for in data science roles. Interactive dashboard with filtering by skill and work type.
Trained on 6,700 real salary records. Random Forest achieves R-squared 0.97. Visitors pick a role and experience level, get an instant estimate with model comparison.
August 2026 to Present
November 2025 to Present
June 2023 to October 2025
January 2018 to June 2024
University of Virginia
Rift Valley University