Justin Verlin

Background

I grew up tinkering — taking apart computers, building small raspberry pi devices, writing little tools when something annoyed me, and building small games for fun. That turned into a real interest in systems: how software gets built, how data moves, how ML models actually make decisions. I'm now a senior at Bucknell studying Computer Science & Engineering and Data Science with a 3.71 technical GPA.

Most of what I build lives at the intersection of backend infrastructure and applied ML: FastAPI services, RAG pipelines, embedding-based retrieval, agentic AI systems, and deep learning models I train from scratch. I care about shipping things that work end-to-end, not just prototypes — right now that means Metrify, a fitness app I took from idea to a live beta on iOS and Android, and AI systems for healthcare workflows at SpinSci.

Education

• Bucknell University, College of Engineering — B.S.E. in Computer Science & Engineering and Data Science (double major), expected May 2027. It's an engineering degree through the College of Engineering, not a straight CS degree. Technical GPA 3.71.
• Teaching Assistant for CS 204 (Data Structures & Algorithms) and CS 205 (Software Engineering & Design), assisting 30+ students.

Highlights

• Metrify — founded and solo-built a full-stack iOS and Android fitness app pairing a chat interface with real tracker accuracy. Agentic AI coach logs food, exercise, and weight through conversation, grounded by RAG over a vectorized nutrition database. In internal beta with ~30 testers.
• Keystroke Synthesizer — trained a DeBERTa-based heteroscedastic regression model that predicts per-character dwell and flight times. End-to-end pipeline: data collection, training loop, FastAPI inference service, and the interactive demo on this site.
• RAG chatbot (this site) — FastAPI + SQLite + sentence-transformers + OpenAI. Indexes my own bio and projects so visitors can ask the site questions directly.
• SpinSci — AI & Product Development Intern since May 2026, continuing part-time through the school year. Built an XGBoost patient no-show classifier on HL7 data (0.745 ROC-AUC) with a "what-if" scheduling sandbox, an LLM call-analysis pipeline on WhisperX + GPT-4o, and a company-wide document management system that turns any file type into searchable markdown.
• Eagle Scout with 3 palms — was an SPL for 6 months, led a 12 man team for 120 hours worth of work on building a fence for a community garden, led a 70 mile week long canoeing expedition, and more.
• Co-founded Bucknell's Fresh Fishermen club and serve as its current president.

Stack

Outside of tech

Warhammer 40k (generic chapter space marines). Some Helldivers 2 when the schedule allows. Hiking & Fishing. Camping with my friends in the summer.