Data Engineer · Analytics Engineer
AI-native engineer building data and analytics systems.
Based in NY.
Experience
dbt Labs · New York, NY
Apr 2026 — Jul 2026Publicis Groupe · New York, NY
Jul 2025 — Apr 2026SUNY Research Foundation · Old Westbury, NY
Jan 2025 — Jul 2025Education
SUNY Old Westbury · Old Westbury, NY
Conferred May 2025Relevant coursework
Capabilities
Languages & Tools
Data Platforms
BI & Visualization
Methodologies
AI, ML & NLP
Spoken
Selected Work
I wanted to see what a city looks like while it's breathing. So this replays every Citi Bike trip in New York from this day one year ago, rebuilt nightly: 172,783 rides across 2,176 stations on the last build, each trail routed between its two docks with the trip's real duration spread along it. Pick a place or a station to see where its riders go, or where they come from.
I wanted a way to learn metrics I hadn't worked with yet, and one source of truth to reference and iterate on. Organizations vary these; this is a place to start. 357 metrics across 11 domains, each with a definition, a MetricFlow metric and a place in a driver tree. Open any one to trace its dbt-style lineage back to the models it reads and out to the metrics it drives. Every metric's SQL runs against zero-row dbt stubs in DuckDB, and the tree is validated on every push.
I track my job search in a Google Sheet, and maintaining it by hand is its own small job — every application typed twice, every rejection on top of that. It is tedious enough that it stops getting done, and a tracker you stopped updating tells you nothing. So this builds the sheet from the mail: one command pulls the new job-application email, classifies it with Claude, and prints a table. Nothing is written until I approve it.
A Python Discord bot built to support and study transfer students in computing. It builds peer networks while collecting anonymized engagement data to inform AI-driven advising. Hosted on GCP with the Google Sheets API for storage, it verifies users by university email, assigns roles, and logs activity behind hashed identifiers. Later deployed at Cal Poly and UT Victoria.
A generative-AI web app that lets educators, parents, and students create customizable K–12 worksheets in seconds. Pick a subject, grade, topic, and format, then refine the output with prompts like "simplify" or "make it more descriptive." Built with Django and GPT-4o, with in-browser DOCX editing, answer key toggling, and a community ecosystem for sharing worksheets.
Classifies IMDb movie reviews as positive or negative using NLP and machine learning. After cleaning and vectorizing text with Word2Vec (Skip-gram), several classifiers were trained and compared. A two-layer Multi-Layer Perceptron over 500-dimensional embeddings performed best, reaching an F1-score of 0.8911.
An end-to-end pipeline tracking tech industry layoffs. Raw CSV data was loaded into an AWS RDS MySQL instance and cleaned in SQL — deduplication, naming normalization, null handling — moving from a staging table to a final table before being visualized as a multi-page interactive Looker report covering trends, company breakdowns, and geographic impact.
An interactive Tableau dashboard on public payroll spending in Suffolk County, NY. Data was cleaned and preprocessed in Python to strip sensitive fields and consolidate job titles. The analysis found that over 25% of total compensation went to the police department — more than every other department combined.
Get in Touch
I'm currently open to data and analytics engineering roles. If you're building something interesting with data — or something that's stopped being interesting because the pipelines keep breaking — I'd like to hear about it.