Projects
Stuff I built. Click a card for details.
2026 · Agentic tool-use · LLM evaluation & calibration
Reason-Eval — Reasoning-Audit Agent
Multi-step Claude agent (hand-rolled tool-use loop, no framework) that reconstructs an argument's premise structure — including unstated premises — flags genuine logical failures without over-flagging ordinary compression, cites evidence, and defers low-confidence calls to a human.
2025 – present · React Native, Supabase, pgvector, Claude API
Journal
Mobile-first AI journaling app: entries embedded via nomic-embed-text, retrieved with hybrid semantic + recency scoring, queried through Claude Sonnet. A Supabase cron autonomously runs weekly synthesis via the Claude API.
Summer 2025 internship · RAG & wearable time series
Regeneron — Data Science & ML
Two-path RAG pipeline (semantic + keyword retrieval over PubMed, ClinicalTrials.gov, UK Biobank) serving 50+ researchers, plus an XGBoost model on wearable time series detecting disease-severity progression at 85%+ accuracy.
Fall 2025 · DREAMT · CNN–Transformer, MLP, classical ML
Sleep Apnea Detection
Course project (Physics 188) engineering physiology features from PSG and Empatica E4 data, training CNN–transformer hybrids, MLP, Random Forest, and ROCKET to flag apnea epochs—with careful handling of class imbalance and subject-level splits.
2025 – present · React, Express, PostgreSQL, Twilio
Roseyflow
Menstrual-cycle mood tracker powered by a Bayesian model trained on the mcPHASES wearable dataset (PhysioNet). Personalizes predictions as users log data and sends cycle insights via Twilio SMS.
Essay · Epistemology & mathematical truth
Philosophy of Mathematics
Deep dive into the philosophical questions that underpin mathematical thought and discovery.