Hybrid data engineer and scientist aiming to make better policy. Building pipelines and collaborating with Claude these days. A soft spot for politics, history, and the taste of adventure.
An MCP server that lets any MCP-compatible AI assistant query the OECD's How's Life? well-being database in plain language, applying the report's own method to every answer, with interactive charts and evals against the published figures.
Full story →
Building and refining a RAG app as a personal assistant as I rediscover a favorite RPG of mine, Runescape. Many iterations, trying different LLMs, and refining prompting strategies.
Full story →As a Data Engineer at dv01, I build an LLM-driven orchestration framework out of reusable Claude Code skill files. It automates the most work-intensive parts of onboarding loan data: data modeling and regression testing, and entity resolution for loans arriving from disparate data sources. I also helped build an agentic, CLI-driven assistant for complex data investigations, backed by an MCP server.
internal work
While consulting for PRAC (a US government anti-fraud agency set up to track pandemic aid fraud) as a data scientist at Elder Research, we built a fraud scoring model for pension claims based on risk metrics in collaboration with PBGC and helped recover $262 million related to improper payments.
Read on pandemicoversight.gov →
A blog post for Elder Research on ML Operations: how to make sense of a newer domain of expertise and where it fits on the modern data science team.
Read on Elder Research →Classifying 185,000 Drugs.com reviews as positive, neutral or negative with LSTMs and a mixed-data model that reads each review alongside its metadata. With Angelo Cozzubo and Noah Selman
Full story →
A suite of static and interactive graphs (Altair, Vega, D3) telling the story of Lebanon's economic and food-insecurity crisis from 2019 onward.
Open web app →20M+ rows of Divvy, CTA, and weather data powering a web app with functioning batch, serving, and speed layers (plus a couple of ML models), built for a Computer Science elective.
Full story →
Analyzing tendencies in party switching among 2020 Peruvian parliamentary candidates, with data viz and a Selenium-based crawler. With Andrei Bartra, Angelo Cozzubo, and Marc Richardson
Full story →
Using government policy measures and national health and economic indicators to forecast Covid-19 cases and deaths across countries. With Diego Diaz and Piyush Tank
Full story →
An interactive map of demographic and trade data across 436 US Congressional Districts and Mexico. Built during my internship at the Embassy of Mexico in the US.
Full story →