★ FLAGSHIP MISSION

Neurelle — Agent Harness over NDA Brain Data

Solo-built agentic research assistant over the dHCP dataset (314,007 files, 966 neonates, 12.7 TB). ~34K lines, 88 tools, 525 tests, data-governance + oversight by construction.

Nascent — AI Oversight in Production

Found an oversight failure in a live agent: 0 of 28 human corrections were retrievable. Shipped the fixes in four waves + honesty/control mitigations. 115 PRs, 953 commits.

Quantelle — Quantum ML Weather Forecasting

Lead on the first real-data test of a published quantum reservoir forecaster — 26.6 yrs of station data, run on IBM's 156-qubit Heron r2 (hardware matched simulation at r = 0.925).

LOCAH.ai — Cited Campus Assistant

Engineering lead: a cited assistant over Laurier's public info, with safety invariants enforced in code — every factual sentence must cite a page it actually retrieved, or it refuses.

Laurier Computing Society — President

Lead a 1,000+ member society: agentic-AI workshops, Meet-the-Professionals panels with industry, and bottie — a Claude-powered Discord bot that turns meeting notes into tracked action items.

Multimodal ML — Risky Financial Decisions

A multimodal model combining skin-conductance signals and market data (PyTorch MLP): 0.96 test accuracy vs ~0.88 for either modality alone, with ablations.

QAOA for Weather Data Assimilation

Co-author — 'QAOA: Mathematical Foundations and Hardware Bottlenecks': warm-start QAOA applied to 4D-EnVar weather data assimilation.

MAPLE — GenAI Platform Security Review

Security review of a government GenAI/RAG platform: 15-item risk register (3 high), reproduced a cross-user file leak (IDOR), and wrote the platform's Software Design Document.