Applied AI Software Engineer at Stingray Digital

Architected a Kubernetes multi-tenant agent platform with isolated per-user execution environments and a custom MCP gateway routing tool calls across 16 internal services; reduced end-to-end agent latency by 9% and eliminated cross-tenant state leakage. Built an LLM API proxy (APISIX + Rust) unifying 7 providers with per-request cost attribution; surfaced $9K/month in previously untracked inference spend across 11 teams, enabling model-level budget enforcement. Refactored production conversational AI layer (Teams chatbot + Electron desktop app, 900 users): decomposed monolithic inference loop into typed pipeline stages, reducing error rate by 35% and p95 latency by 55%.

Software Engineer, Quantitative Research at CDPQ

Built a multi-factor portfolio optimization engine (Python/cvxpy) with risk-parity and mean-CVaR constraints backed by a Snowflake pipeline; reduced tracking error by 15% and solver runtime by 40% via constraint reformulation and warm-starting. Developed a TypeScript dataset processing library consolidating 3 divergent visualization pipelines into a single typed interface; cut cloud costs $13K/year and reduced per-analyst setup time from hours to minutes.

Software Engineer, Intern at CDPQ's Otera Capital

Built a Monte Carlo portfolio simulation tool (Flask + React + PostgreSQL + Airflow) replacing a manual Excel process; compressed per-report analysis time from 15 hours to 3 hours (5× speedup). Automated currency rate updates and cashflow reporting via AWS-deployed Airflow DAGs; reduced report generation time by 40%.

Software Engineering Intern at OneDesk

Implemented a matrix factorization recommender (C++) over user-workflow configuration data; surfaced personalized automation suggestions that reduced setup time by 87%.