Research and production systems spanning LLM evaluation, applied AI ethics, and full-stack data tooling.
Systematization-of-Knowledge research + full-stack dashboard
Compared 8 hallucination-detection methods across a 200-case benchmark. The top method reached 90.5% accuracy — notably, a simpler grounded approach outperformed more complex CRITIC/CoVe pipelines.
Uncertainty-aware navigation prototype
Flags risky drop-off zones instead of treating missing information as safe, weighing incident reports, severity, and recency against map-based risk zones.
2026 National AI Hackathon — Government of Alberta / AGI Ventures Canada
Detects circular public-funding patterns through an interactive dashboard, network graphs, chatbot-driven analysis, and automatic memo generation.
I've spent the last five-plus years as a software engineer, and the last stretch of that immersed in AI/ML — building systems that need to hold up in production, not just in a notebook. These days that means LLM pipelines, retrieval-augmented generation, and the unglamorous plumbing that makes a model actually trustworthy at scale. I'm currently pursuing an M.Eng in Systems Science & Engineering (Interdisciplinary AI) at the University of Ottawa to sharpen the research side of that work.