Building an Orchestration Layer for Agentic Commerce at Loblaws

Why It Matters

Alfred turns complex, multi‑API AI workflows into reusable, secure services, accelerating Loblaws’ ability to deliver scalable, conversational commerce across its nationwide retail network.

Key Takeaways

• Alfred provides reusable orchestration templates for agentic commerce.
• System integrates 50+ enterprise APIs via MCP tool layer.
• Secure, privacy‑first design masks PII and encrypts data.
• Observability handled by Langfuse, Grafana, Prometheus for end‑to‑end tracing.
• Pluggable chat UI enables rapid testing of conversational workflows.

Summary

The talk introduced Alfred, Loblaws’ production‑grade orchestration layer designed to power agentic commerce across its massive retail ecosystem. Built on Google Kubernetes Engine with a FastAPI gateway, Alfred abstracts LLM providers, leverages LangChain‑style execution graphs, and connects to over fifty internal platform APIs through the Model Context Protocol (MCP).\n\nKey challenges addressed include scaling conversational commerce, ensuring privacy and security, and providing a reusable template that lets any of the 500‑plus engineers launch agentic applications quickly. The architecture combines a templating library (Copier), a checkpointing Postgres database, a vector store, and an observability stack (Langfuse, Grafana, Prometheus) to deliver end‑to‑end traceability and PII masking.\n\nDemonstrations highlighted a recipe‑driven workflow where a user asks for a shrimp pasta, the LLM identifies ingredients, and MCP‑backed tools fetch catalog, pricing, and inventory data, presenting results via a pluggable chat UI. The MCP also supplies rich UI components (AGUI, MCUI) that render product lists without custom front‑end code.\n\nBy standardizing orchestration, security, and observability, Alfred enables rapid, reliable deployment of AI‑driven shopping experiences across grocery, pharmacy, loyalty, and fashion domains, positioning Loblaws to scale agentic commerce while maintaining enterprise governance.

Original Description

March 3rd, Computer History Museum CODING AGENTS CONFERENCE, come join us while there are still tickets left.

https://luma.com/codingagents

Thanks to @ProsusGroup for collaborating on the Agents in Production Virtual Conference 2025.

Abstract //

Developing AI agents for shopping is just the first step; the real challenge is reliably running them in production across complex, mission-critical e-commerce systems—a significant MLOps hurdle. In this talk, we'll talk about Alfred, our agentic orchestration layer. Built with tools like Langgraph, LangFuse, LiteLLM, and Google Cloud components, Alfred is the critical piece that coordinates LLMs with our entire e-commerce backend—from search and recommendations to cart management. It handles the complete execution graph, secured tool calling, and prompt workflow. We’ll share our journey in designing a reusable agent architecture that scales across all our digital properties. We’ll discuss the specifics of our tech stack and productionization methodology, including how we leveraged the MCP framework and our existing platform APIs to accelerate development of Alfred.

Bio //

Mefta Sadat is a Staff Software Engineer, ML Platform at Loblaw Digital in the Greater Toronto Area, specializing in building scalable, data and ML-driven distributed systems. He's currently focused on building tooling around agentic systems.

At Loblaw Digital, he spearheaded the Helios Recommendation Engine, a key MLOps solution that personalizes shopping experiences across PC Express, PC Optimum, and other Loblaw apps. He holds a Master of Science in Computer Science from Toronto Metropolitan University.

A Prosus | MLOps Community Production