Startup Mantra: Offering AI platform to grow revenue, boost efficiency and improve customer experiences - Hindustan Times

Startup Mantra: Offering AI platform to grow revenue, boost efficiency and improve customer experiences

BySalil Urunkar
Jun 23, 2023 11:27 PM IST

Instabase, an applied AI tech-startup founded in 2015 by Anant Bhardwaj, is bridging the market gap through its “AI Hub” ecosystem, an innovative repository of AI applications focused on content understanding and a suite of generative AI-based tools

Pune: What do government circulars, legal notices and contracts have in common? Jargon, complex language beyond the understanding of common people. With evolving generative AI (artificial intelligence) and its ability to understand and respond in human language, the problem of processing such complex documents will soon be a matter of past.

Anant Bhardwaj, founder and CEO of Instabase has a master’s degree in computer science at Stanford University. (HT PHOTO)
Anant Bhardwaj, founder and CEO of Instabase has a master’s degree in computer science at Stanford University. (HT PHOTO)

Instabase, an applied AI tech-startup founded in 2015 by Anant Bhardwaj, is bridging the market gap through its “AI Hub” ecosystem, an innovative repository of AI applications focused on content understanding and a suite of generative AI-based tools.

Now catch your favourite game on Crickit. Anytime Anywhere. Find out how

Instabase has recently announced a 370 crore Series C funding round at 17,000 crore valuation.

The Pune connection…

Anant Bhardwaj is an Indian Army kid, a computer scientist, software engineer, and internet entrepreneur. Born in Nalanda, Bihar and currently settled in Pune, Bhardwaj completed his undergraduate degree in computer engineering at the Savitribai Phule Pune University (SPPU) and later pursued a master’s in computer science at Stanford University.

He joined the prestigious Massachusetts Institute of Technology (MIT) for doctorate, where he was co-advised by American computer scientist Sam Madden and professor David Karger.

In 2015, Bhardwaj dropped out of his PhD programme to embark on his journey to found Instabase.

Anant said, “My family moved to Pune in 1999 and I did my schooling from Kendriya Vidyalaya. I wanted to join the Indian Army, but was rejected due to colour blindness. Later I did a job at BMC Software for about two years and moved to the US. I had interned at Oracle and Google, but later realised the cultural difference between universities and education in India and the US. I was focussed more on classes while my friends explored new ideas and started their own companies. I did not know if I really wanted to do a job and so applied for a PhD programme. I told my advisors that I do not know what I want to do, but one of the key goals is to start my company.”

“My advisors supported me and said ‘just go and have fun, but do not do anything stupid’. In three years, I did six different projects. One of the projects was a system architecture called DataHub that argued for a Windows/Mac like operating system design which will provide an interface for end-user data applications, by abstracting infrastructure and other complexities underneath. A lot of people started using it at MIT. I thought this looked good enough to start a company. I did not really know who or why anyone would use it. I moved to California to discuss with some investors to check if they would be interested in funding my idea. In March 2015, I decided to drop out and moved to Bay area. I was lucky enough to secure an initial funding of 30 crore at 164 crore valuation in August 2015 from NEA and Greylock Partners.

Investor experience

Anant was on student visa and hence ineligible to work or start a company in the US, but he had received the funding in his bank account. This created a legal hurdle for him. Recalling the events, Anant said, “I was introduced to the investors by an entrepreneur friend. I had borrowed a bike from another friend on the day of investors meeting, but I reached 45 minutes late due to flat tyre. The investors still took me seriously since I was referred by a credible startup founder. I had no team, no history and still they took a bet on me. I think this is what differentiates the startup ecosystem and investors from India and the US. I had money in my bank, but my visa got rejected. I could not establish an employee-employer relationship with the company since I was my own employer. This turned into a serious issue, but then MIT came to my rescue as they admitted me back for the PhD programme. We hired lawyers and found a solution and later I got my green card.”


Anant said, “We had started with two random apps — Notebook for universities, and Refiner for data scientists to help structure text data coming from sensors and other machine generated sources. We had also built a couple of visualisation apps Lens, and Viz, primarily eye candy for demos. The first 12-18 months were primarily focused on building the foundations of the operating system — a general purpose abstraction for files [file service], a general-purpose abstraction for tabular data [table service], and a general-purpose abstraction for processing tasks [distributed task processing]. None of these had any real value or practical application from the usefulness perspective. It was late 2016, we were a three-person company, in a random demo, a data scientist from a HR and benefits company asked us ‘I want to do some analysis over health insurance plans, would not it be great if we could use Refiner over the insurance plan docs which we get as PDF from different insurance provider websites’. We did not have anything to handle PDF, but we knew how to do extraction from text.”

“We got our first customer, however, the CEO of the company got fired and the deal never happened. We got few other customers who were paying us 25,000 to process their PDFs. Later Standard Chartered came in as our customer and we proposed charging them around 12 crore for their document processing. Since we could not differentiate between customers regarding pricing, we had to let go the smaller customers. Meanwhile, a16z approached us for the Series A round of 188 crore at a valuation of 1,230 crore. So, during this round we had zero-dollar revenue and no product. Later, we sealed our deal with Standard Chartered and another insurance company,” Anant said.

Unicorn status

Instabase grew quickly from 2018 to 2019 where it got five customers and a revenue of about 49 crore. At this stage, Index Ventures and other investors put in 861 crore at a valuation of about 16,000 crore in 2019.

Anant said, “We were a team of 25 people then and we had found a use case where people were willing to pay us money. Thereafter, we struggled a lot to build a sales team and our revenues plateaued. We went from 49 crore to 57 crore between 2019 and 2020. The coronavirus outbreak and subsequent lockdowns gave us time to build our teams and grow. Meanwhile, the market and technology were also evolving. Things turned better as we were getting new customers.

Artificial Intelligence (AI) was evolving, and I had a friend who was in the team which developed conversational AI. Until then we were selling to only large enterprises and our customers were all top banks and insurance companies globally, including four out of five banks of the US. Our revenues grew 6x.”

OpenAI partnership

Elaborating on the agreement with OpenAI, Anant said, “Our customers care about their information privacy and secrecy. While OpenAI stores and learns from data, we entered an arrangement where we will build models on top of their language model, but they will have no right to retain data and no right to learn anything from it. We signed this agreement in February 2023.”

In 2020, Instabase heavily invested in transformer-based layout-aware language models and made several base foundation models available for customers to fine-tune according to their specific requirements.

Anant said, “A key aspect of AI is the ability to learn language. Language understanding was the holy grail of the AI problem. Google’s transformer fundamentally changed the research in AI in 2017-18. The architecture was about can machine learn how to fill any blank or keep adding one blank and keep generating. Earlier, AI models were of few gigabytes and later those became hundreds of gigabytes. Training one model costs around 275 crore and iterations costs were additional. OpenAI was working on this idea and released a conversational engine on top of the language model they had and which later surprised us all. Earlier, we had to hire coders to make machine understand human language and only way to communicate with the machine was through an application which provided a user interface to do something. That fundamentally changed and it was fine for spoken language.”

“For documents it was written language. A document is nothing but artifact of two-dimensional representation of language and we thought what if we figure out a model to represent the two dimensions of the language. By leveraging large language models like GPT-4 and incorporating layout understanding, Instabase’s platform allowed users to create custom solutions for document understanding in a matter of minutes. Instabase AI Hub democratised AI by making it accessible and user-friendly for individuals with both technical and non-technical backgrounds. When ChatGPT was launched we approached them for partnership. In January 2023, we started building document understanding engine on top of GPT and launch this product in February,” Anant said.

Anant and his team launched their AI hub product a fortnight ago. He said, “Earlier our platform was mainly designed for large enterprises and our focus was global market. With AI Hub, India could be a huge market for us. We believe anyone can build an application on top of our AI platform. We support all languages, including regional languages in India. We have offices in New York, Bay area in US, London; and Bengaluru in India.”

Eyeing IPO

Predicting future developments, Anant claims, “AI will change a lot of fundamental things. AI can now help individuals and professionals like lawyers, politicians, teachers, students, etc. We will see a lot of exciting stuff in AI in coming months. Possibly we can build AI agents for commoners who can book restaurants, flight tickers and file taxes. AI is in a nascent stage and we are making sure that lot of people can understand, build, and use a lot of third-party application ecosystem. Eventually our goal is to become a publicly listed company in the next three to four years.”

Share this article
Story Saved
Live Score
Saved Articles
My Reads
Sign out
New Delhi 0C
Thursday, June 13, 2024
Start 14 Days Free Trial Subscribe Now
Follow Us On