Background

Every intelligently run business wants individual purchase data. But linking real world purchsaes to individuals is a considerable challenge that requires Fooda's own POS.

Overcoming the challenge of designing and rolling out a POS was top of mind. Building technology that can process more than a hundred million dollars per year requires world class reliability.

Last, but not least: most of Fooda's volume happens all at once. Turns out most office workers tend to buy lunch at the same time. Successfully handling peak demand requires thoughtful infrastructure and performance decisions.

The Solution

We developed or own internal, highly scalable enterprise grade POS system. Optimized for performance and availability, it's built on a distributed architecture and supports offline transactions. Because we don't believe a bad internet connection should stop customers from buying lunch.

The POS backend tracks every order, event log and user click in real time to provide the best possible user experience.

To keep the line moving, we have a strict requirement on the amount of time to process an order. We use a distributed infrastructure to maintain low latency and high availability. We are built on top of a messaging architecture that links various systems over RabbitMQ. Because you really don't want to hold people up when they're waiting for al pastor. Trust us. You don't.

Background

Historically, the team had an over-reliance on manual QA. As the engineering team grew, the QA team needed to grow in equal proportions. Testing and deployments got slower and it hampered us from iterating quickly.

It was taking one full day of manual regression testing (2 QAs) to sign off outgoing releases. Deploying code to production took up to four hours, with painful manual quality control nad deployment workflow steps.

Because of this overhead, the team could only release 1-2 times per week.

The Solution

The Fooda engineering team loves to solve problems and we're enthusiastic about finding new and improved ways to work together. With that attitude, the team was able to make serious progress with improvements.

The development team strongly believed that test driven development (TDD) was the way to go. During weekly lunch & learns, engineers gave presentations on how to write good tests and how to incorporate TDD into day to day coding.

We migrated the devops stack to Kubernetes running on AWS; created internal tooling to automate testing and deployment steps, and trained engineers how to navigate their own devops issues.

All regression testing has been automated and all new features ship with automated tests. Various levels of unit and integration tests are automatically running on proposed code changes. Today we only need two QAs for twenty engineers and we're deploying multiple times per day in just 15 minutes.

In addition to TDD, we established monitoring and alerting as the cornerstones of the development process. The development team creates health checks, alerting, and dashboards to ensure the health and up time of their own feature sets. Our team finds out about production issues way before a user ever sees them.

Challenge

Before Fooda AI, putting the puzzle together every month was tedious, manual and kept getting harder as the company grew. We needed a smart, automated way to create the best possible restaurant lineup for each workplace - balancing customer preferences, cuisine variety and restaurant availability. We also wanted to provide a sales forecast to the restaurants to help them bring the right mix of menu items each day.