BBC News

Swiftkey, a scientific start-up

Rory Cellan-Jones
Technology correspondent
@BBCRoryCJon Twitter

media captionBen Medlock of Swiftkey talks to Rory Cellan-Jones about his product

Where are you likely to find Britain's brightest young technology firms? In London's Tech City just off the Old Street roundabout perhaps or maybe in Cambridge? So it was a surprise to discover that Swiftkey was housed in bland offices in a dull building in unfashionable Southwark.

But the company started just four years ago by Ben Medlock and Jon Reynolds may have a better chance of making it big than any number of trendy new media Shoreditch start-ups. Swiftkey is an app that makes it easier to enter text on Android smartphones and tablets, and it's already been installed by more than six million users worldwide. It sounds simple, but behind an app that throws up suggested words as you type is some very sophisticated technology and an interesting story of how companies start in Britain.

Ben and Jon met in Cambridge, where Ben was doing a doctorate in natural language processing - a branch of computer science involving machine learning - and Jon was studying physics. In 2008, a year out of university, Jon was working as a civil servant when he noticed how people in meetings were struggling to enter text on smartphones.

"Something leaped out at me - I thought maybe there's a better way of doing this." He realised that Ben's university research might be relevant, and the two got together to make a prototype text entry system.

"It became very clear to me," says Ben, "that it was a perfect opportunity to employ some of the techniques we'd been using at Cambridge in things like recognising language in spam email versus genuine email."

What they came up with was a system that learned language patterns from the user, anticipating what might come next, so that, for instance, when you typed "fancy meeting at the pub for a..." the word "pint" would be suggested even before you typed "p".

But it's a long way from an idea to a product that can pay its way. The duo won a couple of grants from the government's Technology Strategy Board - just enough cash to keep them afloat. "Without the TSB it would have been impossible," says Ben. But they funded the first 18 months of the company's life with just £70,000. "That included filing three patents, building a prototype, and exploiting some old friends to write code for virtually nothing."

The real breakthrough came in 2010 when another government scheme allowed them to visit the Mobile World congress in Barcelona. "We ran around the show trying to demo an early Android prototype to as many people as we could," says Jon. One firm, INQ Mobile was impressed. They said "Hey, we're looking for a keyboard, this is the best we've seen in the show, come and see us next week." Luckily, INQ's offices were in London, not Asia, and the business had its first client.

The whole plan had been to sell to manufacturers, but when Swiftkey launched on the Android Market in the summer of 2010, the app proved an instant hit - despite the fact that they were asking users to pay for it at a time when just about everything was free.

The app was downloaded 50,000 times on the first day and Swiftkey has since built a community of one million active users. They are now looking to the possibility of in-app purchases, customising the service for different customers.

Today, Swiftkey employs 40 people and is recruiting more, many of them Cambridge computer scientists. They are working on refining the technology, which captures the way people use words, and employing it in different languages and for enterprise customers.

For example, in healthcare where doctors and nurses take notes using drug names and a particular style of language, they believe the Swiftkey technology could help staff work more efficiently and safely. "We have statistics that show we're saving people almost 50% of their keystrokes," says Ben "and we're improving accuracy and compliance, which is very important in healthcare."

And his hope is that the work he did at university on artificial intelligence and natural language processing can now find further uses: "These technologies are solving problems that simply couldn't be solved 10 years ago. We've always felt that this had much wider applications - it's just about finding them."

The Swiftkey story is one of lucky breaks, good timing and a bit of a leg up from a couple of government schemes. There is no guarantee that it will continue to grow and prosper in what is a very competitive area. But what is really encouraging is to see two young scientists having the courage and imagination to use their knowledge to build a smart British business.

PS I'm hoping this will be the first in a series of post about promising British technology start-ups. If you've got suggestions of companies that deserve a look, please let me know in the comments below.