Our journey towards solving the language problem for a billion Indians

Sourabh Gupta

Jun 29 · 4 min read

India has 1.3 billion people out of which only 125–150 million understand English. So that leaves us with more than 1.1 billion people for whom internet is largely inaccessible.

If you will take a glance at the mobile internet usage data — there are only about 150 million mobile internet users out of an estimated 370 million in the country today who are familiar with english; the internet is virtually unreachable for more than two thirds of its consumers and that fraction keeps getting bigger everyday.

A billion is a lot of people. A lot of people.

In our case, this simple revelation led to a series of events worthy of an Aaron Sorkin’s movie montage, comprising of sleepless nights spent in remote villages of our country, experimenting with state-of-the-art ML and deep learning tech, and most importantly, forging a bond of trust with people who believed in our hard-work.

Our journey dates back to the freshman year of IITR, when the 3 of us got to know each other through a common campus group dedicated to promote the culture of Entrepreneurship on campus. Quite inevitably, all of us took the decision of not sitting in the campus placements in our final year. It was something that still gives our parents goosebumps but we were certain of what we wanted to do in life, and hence we ended up in Bangalore last year in June as Entrepreneurs in Residence at Kstart.

The initial few months were spent in mostly reading market reports and analyzing potential ideas to work upon. We experimented with the idea of personal savings for a while but taking advice from our mentor and Partner at Kstart, Seed Initiative of Kalaari Capital, Muthiah, we moved on and continued our pursuit of a challenging problem.

We also decided to be more aggressive in our understanding of the masses and spent the next two months traveling to villages and visiting rural areas, talking to farmers, small business owners and people who fell under the category of lower middle-class households. After all of this, we were clear on one fact — There was a huge underlying opportunity among the next 400 million Indians who were about to come online in the next 5 years. And that opportunity was language. We subsequently realized that there is an absence of an interface that can let users interact in their preferred language. We also learnt that businesses are facing a hard time engaging with their multi-lingual customers.

Imagine this: a small Malayalam speaking manufacturer in textile industry, living in a remote village of Kerala, in need of a loan but not enough credit history to qualify for one. At the same time, think of a Maharashtra based NBFC, providing loans to such manufacturers but unable to reach out to people who are beyond their boundary created by Marathi language.

So, even if we remove the barrier of distance, the biggest challenge is that of language mismatch. This opens door for new possibilities in the field of language automation. Especially, for Indic languages like Tamil, Hindi, Telugu, Kannada, Malayalam, Marathi etc. And this is where we come to our idea.

Now the problem that we have on hand here is not something that you can simply cook up after reading a couple of pages on stackoverflow. NLP in indian languages is a hot topic of research in our country. Primarily, in India, there are two famous educational institutes who have done substantial amount of research in language AI (IITB and IIITH). So, to get started we got a couple of professors on board as advisors, Ganesh Ramakrishnan from IIT Bombay and Manoj Chinnakotla from IIIT Hyderabad. Their inclusion was key in helping us understand the language nuances and get started with development.

For the next 6–8 months, we focused on product development. By the end of January this year, we were ready with the basic version of our platform that could build chatbots in Hindi.

To test out our tech, we launched a product called Aisha. Aisha was India’s first personal assistant in Hindi Language for 20–40 year old males living in tier-2 and tier-3 cities. The target segment of Aisha used it for checking their horoscope in the morning, keeping tabs on the cricket scores, getting bollywood gossip and newest jokes to forward on their family whatsapp groups. We achieved significant traction with this product.

Aisha now has more than 50k users and more than a million messages hitting its platform in Devanagari script. This particular product has given us access to immense conversational Hindi data which we are using to improve our NLP engine.

All in all, currently we are working with multiple financial institutions to solve the language problem using AI across different use-cases (lead generation, customer support etc.).

We also recently closed our seed round from Kstart. (Yay!)

Throughout this entire process we met with Muthiah once every month. Each meeting with him was a boost for our confidence and realigned our focus on the critical points.

For us recent grads from college, his words of wisdom were invaluable. Muthiah also connected us to Sumit, who is a partner at Kalaari Capital and a member of our company’s board. Sumit’s deep understanding of AI helped us in quickly reaching to a ‘next level’ of product creation and provided valuable insights into building the company.

I, along with my team, understand that we have a long way to go, for building a company is more like a way of life. Hence, we are always in search of individuals who are not hesitant to take risks and are crazy enough to join us.

A billion Indians are calling. Where are you?