About
My Background
What I build
I like building software products and systems that feel simple to use.
That often means working across boundaries: product definition, backend systems, data flows, infrastructure, frontend surfaces, and the engineering details that make new capabilities dependable.
Problems I like
I'm most energized by ambiguous, high-leverage problems: early ideas, product inflection points, and systems that need to grow without becoming harder to reason about.
Lately, I've been especially interested in AI systems and agentic workflows, especially the work of turning new capabilities into something dependable and useful.
I tend to like work where there are many possible paths forward, and progress depends on taste, iteration, and good feedback.
Writing and outside work
I write about systems, AI, and learning in public. If you want a better sense of how I think, start with my writing on systems, AI, and learning in public or get in touch here.
Outside of work, I spend time playing tennis, reading, hiking, rock climbing, and hanging out with my furmily 🐱.

Technical Focus
- AI systems and software for AI products
- Agentic systems, tooling, and evaluation
- Distributed systems, backend architecture, and reliability
- Full-stack and product engineering
- Cloud infrastructure and developer tooling
- Embedded and hardware-adjacent systems
Credentials
Professional licensure in Ontario since 2017.
Graduate studies in computer engineering and telecommunications.
Undergraduate training in electrical engineering and pure mathematics.
Latest Writing
Dropout as Implicit Bagging
March 7, 2026
Chapter 7 clarified that dropout works so well because it approximates bagging over many thinned networks with shared parameters.
iPad Air M3 11-Inch: First Impressions
March 5, 2026
For study-heavy workflows, a lightweight iPad Air setup with a keyboard, stand, and Pencil Pro can be a practical laptop replacement.
Making Responsive Images Just Work
March 4, 2026
Instead of manually managing -sm/-md/-lg assets, I moved to a manifest-driven workflow that reduced duplication and made performance outcomes more consistent.