Solutions Engineer at Sama — built the company's first agentic AI evaluation offering from zero.
Six years taking AI from concept to deployment across very different domains: computer vision for sustainable aquaculture at ReelData AI (pre-seed → Series A), GenAI at scale in financial services at EY Canada (Tech Lead, 12K calls/day), now agentic AI evaluation and safety.
Comfortable across the full arc — research and engineering through to product, pre-sales, and commercial.
Reading Weapons of Math Destruction in an undergrad philosophy course was where it started — the conviction that ethics has to be structural, not decorative. That's shaped every choice since: B Corp employers, mission-led ventures, the AI4Good Lab at Mila, and why fuzzd exists.
Adversarial fuzzer for MCP servers — tool poisoning, shadowing, and rug-pull attacks. ~91% MCPTox coverage. Deployable as a CI/CD security gate.
Agent observability — intercepts live runs mid-flight, forks from any step, flags behavioural drift. Framework-agnostic.
Turns a memory or feeling into a professional fragrance composition. RAG · custom NER · scent classifiers.


