Billy McGloin
I build machine learning models for hard, real-world problems where uncertainty, time, and imperfect data matter.
My work spans survival analysis, reinforcement learning, and agentic systems. Take a look around.
Featured Work
FrailNet
A variational frailty model for time-to-event prediction applied to interstate conflict. Built on a Cox proportional hazards neural network with uncertainty quantification via variational inference. Pending publication.
Read more →Moon Rover
Reinforcement learning for autonomous lunar rover navigation under uncertainty. Developed with NASA-affiliated researchers and presented at StatConnect 2025.
Read more →Code Analysis Agent
An LLM-powered agentic system that extracts and queries semantic relationships across large codebases using LangGraph, RAG, and Neo4j graph representations.
Read more →Christmas Lights
A computer vision pipeline that maps 500 NeoPixel LEDs in 3D space from multiple camera angles, reducing mapping time from two days to 20 minutes. Runs on a Raspberry Pi 4B.
Read more →Moovies
A full-stack movie analytics dashboard built in JavaScript, exploring IMDb data across genre trends, box office performance, sentiment analysis of scripts, and individual title deep-dives. Built with a team using web scraping, APIs, and zero prior JS experience.
Read more →Lab
A few things I am actively building -- a music project, some political analysis, and a page showing recent reads & reviews. Rough edges included.
See an in-progress project