Wichert Akkerman

An automated personal stylist that could provide detailed clothing recommendations to women, based on their unique body type. Starting as a basic affiliate site, this grew into a complete service for fashion retailers, providing them with a virtual personal stylist called Barbara. She could be used in many ways: integrated in existing eCommerce sites with a trivial 2-line change in HTML templates, via personalized mailings, on in-store kiosks, and even robot assistants: we had proof of concepts running on Pepper and TiKi robots.
I was responsible for multiple aspects of this project, from being part of the product development to architect and lead developer. I designed the architecture, implemented the back end, and built most front ends, with help from a small global team.
Some data on the service:
• Available in seven languages
• Used by over a million women
• Could handle traffic from all customers of a large European retailer with just three servers.
• Custom data pipeline to import feeds with product data from retailers
• Custom input system for labeling clothing items
• Newsletter had unique content for every user, resulting in an opening rate of 40%, dropping to 25% for users who have been subscribed for over two years.

For TNO and the European Agency for Safety and Health at Work (OSHA), I designed and built a platform to help companies identify health and safety risks and determine how to remove those risks.
Though the project initially started with most of the requirements not known. By carefully looking at similar existing systems and talking with domain experts, we determined what the product should be, and evolved it from there. The end result is a success: it has been translated into 12 languages, has multiple active deployments, and has been an official EU tool for over a decade.
All code is open source and hosted on GitHub.

I was responsible for the architecture and implementation of a customer loyalty program targeting SME retailers in the Yangon area. Our goal was to get a better understanding of customer behavior, using a loyalty program to gather data.
This was implemented by providing stores with a simple Android tablet that staff could use to enter sales data, and from which users could scan a QR code to collect points. These points could then later be redeemed for rewards.
During this process, we had to deal with many expected and very unexpected issues, such as very bad 3G/4G network connectivity leading to devices being offline frequently, shop staff not charging tablets, telcos not being able to track and report data usage for SIM cards, and people trying to use tablets as hotspots for personal use. This required us to build extra monitoring tools and improve the training of shop staff.
The system was deployed to 120 stores throughout Yangon before eventually being shut down.