Alex Miles

A PHP and MySQL application for managing warehouse inventory, order picking, and packing processes.

I was part of a two-person team dedicated to building a more efficient replacement system that utilized trolleys for order picking, allowing staff to pick and pack more orders in a day.

Initially, we worked together to define key terms and design processes before dividing the main elements for development. I developed a deterministic method of guiding picking processes with the trolleys, enabling the detection of changing conditions between actions so the picker could take the appropriate action without the out-of-sync and competing sources-of-truth issues we experienced with the previous system. Genericized this approach to be used with simple discrete actions as building blocks to write new processes quickly.

I also built a modular order packing prioritization system to decide the following order to be packed, allowing modules to be easily changed based on requirements at the current time of day.

We delivered the new system on time and saw a roughly 350% improvement in the peak number of orders shipped per day without significantly increasing headcount.

This system was built as a new prominent feature within an existing salon management application. It was built in PHP using Propel 2 and some Symfony components. Inspired by event sourcing architectures, the system fundamentally relies on a single append-only stock ledger table that tracks changes to stock levels and values. Other tables are essentially projections of this data. The project budget didn't allow for a fully-fledged event sourcing system. However, the approach still allowed the salon to reap some of the same benefits, such as quickly rebuilding projection tables when a bug was resolved soon after adoption.

Used aggregate and sortable Propel behaviors to speed up development and manage some projected data automatically. It includes a configurable stock alerts system staff can use to warn of low stock levels and prompt a reorder from suppliers. It also integrates with the existing overall management application's permission system to restrict who can perform certain operations, such as manually adjusting stock levels.

The system tracks over £50,000 worth of stock in production across around 900 different stock levels stored in 25 locations and dramatically reduces the effort required for stock checks.

As a contractor, I worked closely with the tech lead at Tickr, now CIRCA5000, to plan and execute the deployment of a replacement for the API that powers their flagship mobile app. At the time, the app already had thousands of active users, and the nature of the app—financial investments—and the business's aggressive growth strategy meant there was little room for extended downtime or ongoing issues.

The project involved a lot of preliminary testing, planning, and executing a complex, multi-stage process on the day, including dramatic data migration. We also worked with the app team to make changes ensuring forward compatibility with new API responses after the switch and continued backward compatibility should a problem have forced a rollback. Used Ansible Playbooks for most server tasks and a PHP script for the data migration element.

In the event, our planning and preparation proved their value, and the upgrade went very smoothly, completing well within our allotted 3-hour maintenance window, and no rollbacks were necessary.