1) The Founders
Subodh Kumar, Sanjeev Kumar and Kishore Mundra, all from IIT-Kharagpur. Subodh, also an alumnus of IIM-Bangalore, had earlier worked with global corporations Citigroup and Microsoft. Sanjeev has done his master's and doctorate from the University of California, San Diego.
2) The Idea
When Subodh returned to India in 2014 after a decade or so, he immediately realised the need for a system that could recognise regional language speeches and convert them into text. The benefits could be twofold. First, a vast population with limited or no knowledge of English would find it easy to communicate over phones and messaging platforms. Second, it would help people who are not conversant with smartphones (typing in the vernacular is cumbersome even on smartphones with Indic keyboards). "My parents have smartphones, but they cannot text me," says Subodh, explaining the trigger. All these made him research more about the speech recognition technology and the trio set up Liv Artificial Intelligence to address the pain points.
3) Technology & Growth
After two years of extensive R&D, Liv.ai came out with APIs (application programming interfaces) that can recognise nine major Indian languages as well as English and convert the same from speech to text. The technology is currently used across segments, inclu-ding e-commerce, BFSI, consumer devices and apps, speech analytics and robotics. Liv.ai is also working on text-to-speech conversion, also known as speech synthesis, and voice-interactive devices/apps. The company says it is getting about two million API hits per day and wants to scale it to 100 million by the end of this year.
4) Challenges
Tech behemoths such as Amazon, Apple and Microsoft are also betting on this technology, but they mostly focus on English. If they try to tap the vernacular market in a big way - Amazon's Alexa is already getting an Indian makeover and Siri could be next - it will pose a tough challenge. Plus, it will be difficult to collect the huge amount of data (think of all the language and phonetics variations) which must be incorporated to make the technology as accurate as possible.