Research, analysis & things I'm building
Selected projects across data, business intelligence, and product development. The question behind the work, and what came out of it.
01 / Research
WPI · Team capstone
Spring 2026
What happens when hospitals optimize for the score? An analysis of federal penalty programs across 2,833 U.S. acute care hospitals, with an interactive explorer that makes the research accessible.
Python / statsmodels / OLS Regression / Pandas / Chart.js / HTML/CSS/JS
02 / Product
Independent project
In development
From a spreadsheet to a clearer picture. I’m building a no-code workspace for exploratory analysis, data quality checks, and machine learning, so understanding a dataset doesn’t have to start with writing code.
Next.js / FastAPI / Python / Supabase / Docker / Vercel
03 / Interactive AI
Independent experiment
2026
A diary that writes back. Write with Apple Pencil, watch the ink dissolve into the page, and receive a response in animated handwriting. Separate vision and language requests keep the handwriting recognition focused while the diary maintains its voice.
SwiftUI / PencilKit / NVIDIA API / Node.js / Express
Machine Learning
End-to-end ML pipeline for income prediction using boosted decision trees on Azure ML Studio. Includes data preprocessing, feature engineering, hyperparameter tuning, and model deployment. Achieved strong AUC-ROC on the Adult UCI dataset.
Python / Azure ML Studio / Boosted Decision Trees / Feature Engineering
Machine Learning
Self-hosted voice synthesis backend using Coqui XTTS v2. Clones and consistently reproduces an assistant-style voice from a single short WAV reference clip; no training required. Production-ready REST API with automatic GPU/CPU detection and cached conditioning latents for low-latency inference.
Python / FastAPI / Coqui XTTS v2 / PyTorch / REST API / Git LFS
Developer Tools
A shared storage API for Google Drive, Nextcloud, S3-compatible services, and local storage. TypeScript and Python SDKs let applications work with storage their users already own.
TypeScript / Python / OAuth / S3 / WebDAV
Data Quality
A Python library for auditing messy DataFrames, prioritizing issues, applying fixes, and comparing the results. Includes five correlation methods, a command-line interface, and HTML reports.
Python / Pandas / CLI / HTML Reports
Data Engineering
An air-quality pipeline that ingests WAQI readings, transforms them through dbt models, and flags pollution spikes with statistical anomaly detection. A browser dashboard brings city trends, maps, and rankings together.
Python / PostgreSQL / dbt / FastAPI / Airflow / Chart.js
Research
An experiment following language models across five generations of self-distillation. Compares two teacher lineages and tracks changes in vocabulary, repetition, coherence, and similarity to the original responses.
Python / PyTorch / OPT-1.3B / Model Evaluation
Developer Tools
A FUSE layer that gives an exFAT drive the POSIX behaviour Linux tooling expects, so PostgreSQL, Redis, Docker volumes, git and SSH keys can live on an external disk without reformatting it.
Rust / FUSE / Linux / systemd / Docker
Web
A browser chess board rendered in 3D with crystal pieces, played against Stockfish or another person over a private real-time room. Includes voice controls, puzzles, replays and sound.
TypeScript / React Three Fiber / Three.js / Stockfish / Vite / WebGL
Have a question, a project, or something worth exploring?