I build and evaluate systems where people and large language models work together: a private AI notebook for Mac, a deployed advising chatbot, and a measured study of what a language model can and cannot read in an earnings release.

I am interested in the gap between what AI appears to do and what it measurably does for the person using it. My projects follow the same discipline: define the task, measure the human baseline, measure the model on the same records, and price the result. Outside school I operate a systematic equity trading system I built myself, documented in the Quant Trading Lab below.

• Ember, a private AI notebook for Mac that I designed and built, public since October 2026. Notes are plain Markdown files in a folder on your Mac. The AI you sign in with (Claude, ChatGPT or Google, or the free Ollama app on your own Mac) reads the notes you hand it and writes there. Ember has no ads and does not track you. Website · GitHub page and download · Privacy
• INFO 367 final project, December 2026: the earnings-release study below at 500 or more records, with confidence intervals, a timed human baseline and a cost model against an analyst's wage.

• Aiming to replace the analyst? Can a language model read an earnings press release and call the stock's move? Two careful readers matched the market 33% of the time; Gemini reached 58% at half a cent a release, with a re-run showing the gaps between settings are noise. The page explains why the information is not on the page for anyone, where AI does replace the reading, and what to measure before automating. Team lead, INFO 367, Applied Large Language Models for Business.
• Makena A RAG advising chatbot deployed for Montclair State University's Feliciano School of Business. Project manager and lead developer: an asynchronous Python scraping pipeline, isolated data silos that stopped undergraduate and graduate answers from bleeding together, and a system prompt hardened against more than 35 red-team prompts. Benchmarked at 9.2 out of 10. Live demo.
• Quant Trading Lab A live, fully automated intraday trading system for US equities, designed, built and operated end to end, with the research and risk tooling around it. The engineering is documented; the strategy is private.
• Data Jobs Dashboard An interactive Power BI dashboard over 479,000 global job postings, built with Power Query and DAX: KPIs and salary comparisons by title and region.

Python (pandas, NumPy, scikit-learn) · SQL · Power BI · Tableau · Excel · Bloomberg Terminal · Claude and Gemini APIs · retrieval and evaluation workflows for language models · Git and GitHub

LinkedIn · Portfolio · Email

Open to opportunities in data analytics, applied AI and quantitative research.