Building Agents at Shopify | In-person

Join us for MScAC Talks, a yearly speaker series where academic and industry leaders share applied research driving real-world advances at the University of Toronto.

Abstract

This talk will provide a technical look at how Shopify designs and evaluates agentic systems at scale. Topics include: how agents interpret merchant intent; how tool‑use and action orchestration are implemented; how the platform integrates with Shopify’s broader ecosystem of APIs such as GraphQL; and how we evaluate assistant quality in domains where correctness, safety, and business impact all matter. The talk will also highlight challenges with building and maintaining agents with several tools and subagents from simpler chat agents to deep research in an agentic graph.

Shopify is redesigning our platforms to be AI-centric for merchants and shoppers. We’re building a new generation of AI‑powered commerce assistants designed to empower entrepreneurs to run their businesses with greater speed, clarity, and confidence in a common platform for all of our agents. At the core of this effort is Shopify Sidekick, an agentic AI system that can understand a merchant’s store, suggest actions, and collaborate with them across tasks such as product creation, store setup, analytics, content generation, and workflow automation. Sidekick is built on a platform that combines large language models with structured commerce data and tool‑calling, enabling it to operate reliably for millions of real merchants.

Attendees will gain insight into the engineering patterns, developer tools, product considerations, and research challenges involved in building practical, domain‑specific AI assistants—and what this means for the future of applied AI in commerce.

Speaker Biography

Justin Harris has spent over a decade building AI assistants at planet scale. As Senior Staff Engineer at Shopify, he is the tech lead for design and architecture of the agentic platform that powers many of Shopify's AI assistants. Previously at Microsoft, he architected core components of Bing Chat and Microsoft Copilot, designing a patent-pending extension framework that serves millions of users across Edge, Windows, and M365. Earlier at Maluuba (acquired by Microsoft), he built deep learning NLP systems for conversational AI. He holds a BMath in Computer Science and Combinatorics & Optimization from the University of Waterloo.

About the MSc in Applied Computing Program

Housed in the Department of Computer Science at the University of Toronto, the Master of Science in Applied Computing (MScAC) program is committed to educating the next generation of world-class innovators. The program bridges cutting-edge research with industry innovation, equipping students to translate ideas into real-world solutions. By fostering collaboration among students, faculty, alumni, and industry partners, MScAC creates a vibrant community where research excellence drives meaningful impact.

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher Master of Science in Applied Computing @ UofT