Learning from Our Natural Language and Voice Search Internal Hackathon

Voice is a huge topic for Algolia. The best perspective I can provide is that in early 2018, we had one or two customers each week asking us about voice in passing. Today, most of our conversations touch on voice. We’re at an exciting spot where our core interests are meeting the market demands perfectly.

Voice touches on everything

Everybody has a hand in the future of voice. We need to provide good relevancy for natural language queries, build services that enable people to adjust that relevancy on their own, and create tools that help developers implement voice search on any platform. We also want to work directly with our business users to design the best voice experience This isn’t the effort of a small group of people, but of an entire company. All of Algolia believes in the future of voice search.

That’s why I loved this recent internal hackathon we held. We invited the entire team to participate for two days to experiment with natural language interactions. We wanted everyone to daydream about where voice could head.

A hackathon for everybody

Often, when people hear “hackathon,” they think of groups of devs huddled in a circle, click-clacking on keyboards. It’s true that, in my opinion, our engineering team is one of the best for our size in the world. But we also have talented people in every single job role. If we limited the hackathon just to the coders, we would lose out on these other perspectives. While we called the event a hackathon, we also thought of it as less of an event for hacking and more of an event for building. Anyone can build, even if they can’t code.

I think the team that best embraced the call for unique viewpoints was “Team 28.” When you went by their sprawling area during the event, you might have wondered if their name represented the number of team members they had. (In truth, Team 28 had eleven people.) Software Engineer Jonathan Montané started the team, and he was intent on recruiting from as many different job functions as he could.

Montané told me that the mixture of coders and non-coders was vital to their project’s success. He said that because voice app builders spend much of their time understanding how users will interact with the apps, the non-technical members of the team were able to help decide what features Team 28’s app would support. This led to the creation of “primary” and “secondary” functionality that the coders on the team might not have discovered if they had started immediately with the code.

Algolia’s an international company. We have six offices around the world, and even in our Paris office nearly a third of our employees come from outside of France. Everyone brings a unique perspective.

We wanted all of our offices to participate, and I was happy to see our New York and Atlanta offices get involved. Nearly every person in these offices contributed–incredible when you realize that these are primarily sales offices! New York focused on a common Big Apple problem of what to eat for lunch, with an assistant that first determined the identity of the asker before turning to Algolia search to provide restaurant recommendations. Atlanta, meanwhile, built an Alexa skill to serve their own sales needs. A user can ask for “opportunities in pipeline,” and Algolia’s query rules will determine intent to provide the right answer.

Whats next for Algolia?

The hackathon projects have helped us as we expand our voice search and natural language processing capabilities. Don’t fret–we aren’t shipping code directly from the hackathon. We did, however, learn a lot about what developers with little NLP experience are looking for as they work with existing tools. We’re now taking this experience and combining it with customer feedback and our understanding of where voice is headed in the coming years. There is much opportunity for natural language search, and I expect that Algolia is going to continue our leadership.

I hope people will check out what we have to offer for their voice search needs on mobile, voice-first, or anywhere else. Check out what we’re doing at algolia.com/voice and let’s talk.