Experience
Meuze builds AI software that helps restaurant chains run their stores, and its system is in production at multinational food and beverage chains. I worked across the TypeScript monorepo (Next.js, NestJS, Supabase) and the Python agents that run on AWS Lambda. My main work was the forecasting agents, built with LangGraph, which combine time series models (TimesFM, Chronos) with business rules to recommend what each store should order. I also worked on the pipeline that evaluates and retrains them, and I designed how services authenticate to each other, with short-lived JWTs issued by the API and verified by each agent.
Empor
Empor is a student marketplace in Montréal with about 1,200 users. In the React Native app I built the product pages and image galleries, and brought them from about 35 to 60 frames per second with caching and memoization. On the Spring Boot backend I tracked down why the database connection pool kept running out, which came down to JPA lazy loading and the HikariCP settings. I also added role-based access control and wrote integration tests with Testcontainers.
101.org
I built and maintained MERN web applications (MongoDB, Express, React, Node) used by about 500 people, including their APIs, secured with JWTs, and a move to a sharded MongoDB cluster.
Activities
DIRO podcast, Université de Montréal
I edited and produced episodes, and interviewed researchers, among them Gilles Brassard.
McHacks 12, McGill
We built Patiently, a hospital dashboard that predicts wait times with a PyTorch model. It used React and Flask, with Auth0 login and live updates over WebSockets, and it won the MLH prize for Best Use of Terraform.