Tech in Asia peeked inside The Hive – the place where the coders work – and caught a glimpse of what goes on within. Here’s what we learned
In the heart of the Sandcrawler – a building modeled after the huge mobile fortresses in Star Wars – sits an army of coders who are using technology to improve how the Singapore government does things.
They’re part of the Government Digital Service (GDS), a unit started initially in secret and without permission, but has ballooned to a team of over 100 hackers.
Tech in Asia peeked inside The Hive – the place where the coders work – and caught a glimpse of what goes on within. Here’s what we learned.
They’re run like a startup
When you enter The Hive, you wouldn’t think it’s a government unit at all. There are standing desks, the office is open, and you hear words like agile methodology bandied about.
Agile is something the team practices. It’s a philosophy of software development which emphasizes developing workable software frequently and allowing for adjustments on-the-fly.
While government websites typically wait months before their first deployment, the team aims for one deployment a month, along with small bug fixes in between.
Continuous integration is another philosophy the team adheres to. That involves developers integrating code into a common repository a few times a day. Tests are then done automatically on the new code to detect errors.
This practice purportedly detects and locates errors more quickly and easily.
They’re building a major site on React.js and Ruby on Rails.
The bulk of GDS is focusing on building the Business Grants Portal, a website consolidating various government business grants into one destination so people can easily apply for them.
Data.gov.sg is built on Amazon Web Services.
Government sites with confidential data are hosted on a private cloud. But Data.gov.sg, a repository of public government data, is hosted on Amazon Web Services, the cloud provider of choice for many commercial apps.
GDS blogs about its approaches and processes – a lot.
Like many startups, GDS has been constantly explaining how it does things. It runs its ownMedium blog, in which it dives deep into the technical aspects of its work.
They built a tool to test apps simultaneously on iOS, Android, and the web.
It looks like something out of The Matrix, and it’s designed to speed up and automate testing.
It’s customized to test a couple of apps they’ve built, but can be tweaked to test new ones. The tool has a suitably geeky name too. It’s called Rapsel, short for Robot framework with Appium and Selenium. Robot, Appium, and Selenium are open source software testing tools.
They’ve dedicated rooms for usability testing.
Software creation isn’t complete if users aren’t involved in the process. The Hive has two rooms dedicated to user experience tests.
There’s one where users can try out apps as part of a focus group, and another contains devices to track a user’s eyes as it moves across a screen.
“We try to test at every phase, from wireframes to the actual website,” says product developer and manager Yang Zi Dong. She’ll bring in actual users, such as owners of the companies who apply for grants on the portal, to try out new features.
The team’s pretty diverse.
GDS has a couple of engineers who’ve worked in ST Electronics, a government-linked firm that builds electronics and communications systems. It has an engineer who was a lead developer at a gaming company. It has software developers, data scientists, geospatial engineers, UI/UX designers, and internet of things engineers.
They do a lot of experimental prototypes
Besides big projects like the Business Grants Portal, a lot of experimentation happens at the Hive too. Various government agencies would come up with problems, pass them to the GDS, who would develop solutions.
For example, Jurong Town Corporation wanted a cheap but accurate way to survey old factories to know what repairs were needed. So the team used a drone to fly around a factory, scan it, and develop a 3D model. Inspections were then done on a computer.
Wayne Tan, the geospatial engineer who’s involved in this project, reckons that this new approach is about four times cheaper than having a person survey the buildings manually. It cut costs down from US$40,000 a building to about US$10,000.
Celine Chia, who works in the internet of things group, built a gas sensor network to track how smelly rubbish chutes would get, and let cleaners know if it’s time to clear the chute.
That meant heading down to rubbish sites, placing the sensors, testing them, and integrating them with software. That, of course, brings unpleasantries.
“We tried not to breathe,” she said. All in a day’s work.
This post was originally published on Tech in Asia.