AWS needed a compelling demonstration of computer vision and machine learning capabilities for retail applications at Sydney Summit 2019. The goal was to showcase how physical retail stores could leverage AWS services for real-time inventory tracking and customer engagement through an interactive, hands-on experience for 12,000 expected attendees.

As part of a 3-person team, I built the processing pipeline, AWS infrastructure, and CI/CD automation for the “Retail of the Future” exhibit featuring a replica Amazon Go store called “Smart Shelves.” Working alongside Josh Passenger (app dev, front-end) and Charles Prosper (AI/ML specialist, data prep), we had 6 weeks from initial ask to delivery—requiring 2 major pivots when the original plan to reuse existing assets fell through, and the scope expanded 2 weeks before the event to include 3 additional retail demos featuring AWS Personalize and AWS Pinpoint. These are some of the original product images we used in our model training:

Product Product Product Product

The technical architecture combined dual cameras feeding images (2 per second) to edge compute, which sent data to the cloud for processing. The pipeline used SageMaker for model training, data preparation, and managed endpoint hosting, with EC2 + VPC + ELB handling application hosting. We implemented one-shot multi-feature recognition that returned bounding boxes and labels for all identified products, then visualized the results using Amazon Sumerian to create a digital twin showing real-time inventory and product positioning on the shelves.

System Architecture

This is us testing the impact of camera location on image capture and inference accuracy:

Shelf Build

The Smart Shelves exhibit achieved 100% uptime throughout the event with accuracy exceeding 85% and real-time feedback. Positioned front and center at the booth, most of the 12,000 attendees interacted with the demonstration. The exhibit successfully demonstrated AWS’s computer vision capabilities in a tangible way—visitors could see immediate results as they moved products, making the technology accessible and understandable. Two major Australian supermarket brands engaged on the day and followed up post-event, validating the business impact of the demonstration and establishing AWS’s credibility in retail innovation.