For two days in March, developers around the country will congregate online at a hackathon. Certainly par for the course for Indian tech, but what makes this particular hackathon different is that it calls developers to build some cutting-edge solutions in conversational AI in multiple Indian languages. Aptly called BuildForBharat, this hackathon is being organized by the Bangalore-based conversational AI solutions provider Dheeyantra Labs.

Vidhu Bennie Tholath, cofounder & CEO of Dheeyantra Labs decided that the time is ripe to mobilise an ecosystem that would build for Bharat – a collective term referring to the 650+ million Indians that are rapidly contributing to the next phase of growth in Indian IT services and technology solutions. An integral aspect of building for Bharat is to provide the language support for this section of India’s population.

India’s wildly successful Internet story that began in the 1990s has largely been dominated by the English-speaking and educated urban folk that reside in some of the biggest cities. In the past few years, the Internet has been getting democratized, most notably by the strides made by telecom companies to provide broadband services beyond big cities, largescale manufacturing of smartphones and their consequent distribution, growing Internet literacy and the robust growth of Jio by Reliance Industries. India is the second fastest smartphone market in the world, and is poised to add more in the coming years. Last year, for the first time, the number of rural users of the Internet in India exceeded urban ones. And this growth story is nowhere close to slowing down - the Union government has promised broadband access to all Indian villages by 2022 under the National Broadband Mission, increasing tower density to 1 per 1,000 by 2024.

While the hardware is making its way into the hands of the aam aadmi, the software is playing catch up. And this is where Tholath would like to make a difference. A starting step would be to mobilise an ecosystem whose sole purpose is to think for, and build for this massive population that is waiting to reap the benefits of modern digitization. As I spoke to Tholath, all I could think of was the old adage – if the mountain won’t come to Mohammed, then Mohammed must go to the mountain. Building up rural literacy to levels that the worldwide web demands is a very tall ask, and one that will take an indefinite amount of time – so why not make technology a lot more accessible and palatable to the tastes of rural India?

This analogy hit closer to home for Tholath. He spent more than a decade working for Samsung Electronics, and within that tenure was a sizeable time spent in the company’s South Korean headquarters in Seoul. Tholath and his team were the ones that developed Bixby, Samsung’s virtual assistant. What struck Tholath was how utterly accessible Bixby was to the native South Koreans. And one of the reasons this was so – Bixby was available in the Korean language. “I would see elderly folk, many semi-educated or even uneducated, use Bixby with ease for a variety of purposes like home automation and shopping. Language is a huge proliferator of technology.”

Ironically, back in his native state of Kerala, Tholath’s elderly father who is educated and speaks English but isn’t entirely comfortable with using a smartphone – would call his son in South Korea to seek help recharging his phone or buy groceries online. “This stark discrepancy in access to technology was one of the main reasons I started Dheeyantra Labs.” Along with Sreekumar Jairaj, a friend and another experienced professional in building conversational AI systems (Jairaj was the R&D Manager at IPSoft, and was part of the core team that built the company’s conversational AI platform Amelia), Tholath founded Dheeyantra Labs in 2017.

Initial months of launch saw the duo conduct a detailed market research exercise, which revealed conversational AI was a growing need in business - what limited its growth was a tainted perception of the technology’s true capabilities. “Early days of chatbots were raw and didn’t really help businesses like they are today. It made early adopters question the cost of production and R&D. One of the significant impediments was the machine’s inability to truly understand customer needs – and today’s bots are successfully tackling this issue by decoding language.”

So the second order of things for Tholath and Jairaj was to build a truly smart, intuitive machine that understood languages. This included hiring a number of linguists in addition to developers, to work on making the machine understand semantics of local Indian languages. Dheeyantra Labs has filed five patents in various applications of NLP and NLU. The AI’s language accuracy for Hindi and Malayalam is at 98%, while Tamil is at 91%. Unlike several other language AI solutions, Dheeyantra’s machines are trained to understand a range of Indic languages, inflection, dialect, tone, meaning and more – making the machine truly understand a language, instead of relying on an initial layer of machine-driven translation. One of the areas that Tholath’s team is working on is Indic OCR – or Indic Optical Character Recognition. The solution currently supports nine Indian languages, and can help users decode content in a language they choose. The app for this solution called Äkshar is currently in the pipeline.

Another key offering is Aham, which Tholath refers to as an “alter ego”, a bot of bots, a botfather even – an all-encompassing bot that seamlessly integrates crucial data points for the benefit of the user, and available in multiple languages. In the early days of COVID19, Aham collated data from WHO and the MoHFW to dispense accurate information about the pandemic and related data, in voice and text. Now, the platform is being extended to manage pandemic response better - Dheeyantra Labs is working with several NGOs as part of an eight-member industry consortium to assist healthcare workers on the ground. Soon, it will be modified to assist with the national vaccine rollout.

Tholath believes there are some exciting days ahead for conversational AI in India, and regional languages will witness a resurgence through technology. The upcoming hackathon is the first step towards this.

Dheeyantra Labs’ BuildforBharat hackathon will take place from 26th to 28th March online. Registrations close on March 10th. For more details, visit https://www.dhee.ai/category/dhee-hackathon-2021/