A Bot Whisperer’s Pursuit of Intelligence

Tar One is convinced about his mission as he exits the Google office in Bangalore one last time: build a platform for users to meet new people online… and rate one another after. It’s his great idea to ensure people behave nice.

Yet, anyone who’s watched the Black Mirror episode Nosedive would know how terrifying that can be. It has people rating each other after every in-person interaction. The ratings dictate their social standing, their friends, where they live, where they work, and what privileges they are entitled to. So the people in it behave excessively nice. But when a young woman has an outburst at the airport, her rating plummets and her life spirals into ruin.

That saccharine but dystopian world isn’t Tar One’s mission. Most social platforms are primarily about recreating existing networks. Tar One, who had learned to quickly adapt to new environments growing up in India, genuinely wants people to be able to make new friends. His Nice Point System for rating people works smoothly enough that some users end up marrying their newfound friends. His eyes light up as he says this.

Tar One is Tarkeshwar Thakur, the son of a schoolteacher from a village dotting the Indo-Nepal border. He would assume his bot personality a few years into creating his platform, Chatimity, a portmanteau of chat and proximity.

Chatimity transforms Tark from an IIT-bred and Google-nurtured 10x engineer into a well-rounded entrepreneur. It nudges him toward his calling even after he’s sold the platform to Freshworks Inc. At the cloud software company, Tark helms deep-tech projects at the frontiers of sales, support, and customer engagement.

Among these is a repository to pool the more than 50 TB of data generated every month across all Freshworks products to make them accessible to everyone; a critical technology hack for chatbots that Freshworks founder Girish Mathrubootham calls the Holy Grail of customer support; and a smart bot system that can detect likely intent and hold intelligent conversations.

It’s probably premature and even unfair to state these to be Tark’s calling. Because, accomplished as he is, Tark seems aware that he’s left a few big tasks unfinished and a few large challenges unresolved.

The making of Tark

Murli-Parsauni is a tiny village in Bihar state’s West Champaran district, where a farmer’s uprising triggered India’s first Satyagraha movement for freedom from the British. Fewer than a thousand families occupy the village. It has a middling literacy rate and most of the villagers are engaged in agriculture or run local businesses. Tark was born here in 1978.

Good for him that his father, Jagdish Thakur, was a schoolteacher and among the most educated in that rural nook. “He taught us Shakespeare when we were in third or fourth standard, even though we didn’t understand the significance of it at that time. In an extremely remote area where we got English newspapers a day late, my father would insist on us studying Maths and English. That kind of set us apart,” says Tark, reminiscing about his childhood from Freshworks’ modernistic office in Chennai, more than 2,000 km from his birth home.

His life first changed course when Tark was selected to join Jawahar Navodaya Vidyalaya, a federal government-run alternative schooling system for talented children. That was in grade six.

A few years later came another life-changing moment. For ninth grade, under an inter-JNV exchange program, Tark was transferred to the Nizamasagar school branch some 1,500 km away. “That year, they were sending mostly troublemakers from the school. My name came up in a lottery for the three remaining spots.”

It’s hard to imagine the affable, mild-built and spectacled Tarkeshwar Thakur in dad jeans and plain blue t-shirt as a troublemaker. Whatever the reason for his ‘migration,’ it proved transformational for him.

Tark topped grade 10 in Nizamsagar, returned to West Champaran an elite student, became school leader, topped grade 12 as well, and got selected to the premier Indian Institute of Technology in New Delhi. There, he ran into sobering reality. After the heady rush of high school, Tark found himself in an alpha environment of school toppers and city-breds. “It was a great leveler. It toughened me.”

Becoming Tar One

Much of what Tark is obsessing over now at Freshworks can be traced to this phase that saw him evolve beyond his formative horizons. He began reading more widely, experimenting with the stock markets, diving into programming competitions, working with global tech giants, and starting up.

After a brief stint in Silicon Valley, Tark returned to India when one of his two younger sisters fell ill. In pursuit of his father’s dreams, he flirted with the idea of becoming a bureaucrat for a brief while. After another stint with a startup near Delhi, he landed in Bangalore to join Oracle, his first tech giant. In under a year-and-a-half, he jumped to Google.

There, Tark initially got involved with the Maps project. He then moved to social search, which involved pulling data from Orkut to surface results, and later to music search—both critical projects and highlights of Tark’s five years at Google.

Tark also began reading annual reports and trying to make sense of how businesses are run, and imbibing a Google-derived engineering worldview. “Google had a big influence on me,” he says. “But I got a sense that maybe I should try out something on my own.”

Tark founded Chatimity in 2011 shortly after he left Google. His cofounder, Aravind Murthy, also from Google, was yet to join. After about a month building the app, Tark one night decided to upload it on Google Play Store. When he woke up, he found users active on the app. “They were all liking it. They were saying good things about Nice Point System. By the end of the day we had 200-300 users.”

Nice Point System—the well-intentioned people rating mechanism—was a core differentiator for the app. An attempt at influencing good behavior online and in person without overt policing. A few other similar efforts at creating a Yelp for people appeared a few years later. The Black Mirror episode aired about half a decade later.

A couple of years down, Tark and his small band of engineers built a bot to engage the growing number of users from across the world, including the US and Europe. A lot of effort went into giving Miti an appealing personality and keeping it impartial.

“While at Chatimity, we had a visitor from Australia, a user of the app,” Tark tells the audience at a Freshworks SaaS@Scale event in Bangalore. “On the app, we were all known by our nicknames. I was Tar One. She met all of us and then says, ‘Where is Miti? He replies to my messages very quickly.’ She hadn’t realized Miti was a bot.”

From the front row, you could sense the pride.

Mayank Dixit, who was the sole frontend developer at Chatimity, says Tark carried with him a Google mindset—move fast, break things. “He was a people-first guy with great energy. As a mentor, he would guide you and trust you entirely.”

Chatimity grew to about 3 million installs and 30,000 daily active users and was making money—but the hockey stick remained elusive. So the team began building chatbot prototypes to sell to some of India’s biggest internet companies. “Many of these quickly turned into acquisition discussions. We didn’t go with that intention. But I also realized we would have a better product-market fit in B2B,” says Tark.

Tark also approached Freshworks. “We soon had a discussion and Girish made an offer on the first day itself. He has a strong gut feel.”

In October 2016, Freshworks acquired Chatimity for an undisclosed sum.

Quest for intelligence

At Freshworks’ office in Chennai, Tark began work on improving production performance in Freshdesk, the company’s flagship helpdesk product. Shortly after that, he was leading a team to build a data lake called Baikal, named after the world’s largest freshwater lake by volume. Alongside, he started work on Freshsales, the sales CRM software, and filed a patent for the smart lead-scoring system his team developed.

Then sometime in February 2019, in response to a Workplace discussion on the top problems to solve for Freshdesk, someone brought up “the eternal helpdesk problem—can Freshdesk not reopen tickets where end-users get back with a ‘thank you’?” Girish tagged Tark.

In the customer support world, that universal expression of gratitude and its myriad versions are pesky bugs. An “awesome” or a “works now” could end up reopening a help request, complicating matters for customer support teams.

Barely a week later, Tark posted a prototype of the Freddy Thank You Detector on Workplace, tagging his team and thanking Girish for the ‘Holy Grail’ idea.

It was a huge hit. The CEO was thrilled. “This is awesome. Our competition is still asking customers to reply above the line,” Girish wrote, punctuating his reply with smileys and applause emojis.

Tark is particularly proud of this achievement. “We developed a state-of-the-art model that captures a lot of the nuances of language,” says Tark, now vice president of engineering at Freshworks. “That was the first deep-learning model we deployed in Freshworks.”

Tark’s most crucial work at Freshworks has been on the company’s artificial intelligence project, represented by the cuddly and bespectacled puppy, Freddy. His latest effort is a bot feature capable of understanding a customer’s likely intent and requirements and responding intelligently even while conversing in a free-form way.

Tark with some of his colleagues from Chatimity shortly after Freshworks acquired the company.To Tark, Freddy represented an expressway to the future.

“In the original presentation I made to the team on why we were doing the Freddy project, I had an image of a boat charting ahead in the big blue ocean. The idea was that now that we had caught up with our competitors in features, we needed a Blue Ocean strategy for AI. We needed to differentiate ourselves and be ready for the next waves to come.”

For all his achievements, Tark realizes he has a long way to go, a lot more problems to solve. He understands that the world of sales and support faces imminent upheaval from AI and the use of newer machine-learning and deep-learning techniques. And he’s anticipating it with the excitement of a school kid who cracked his first entrance test to a premier school, wrote his first code.

“There are several interesting problems for which you may not know the answers, but you know how to arrive at them. When even those pathways are not known, navigating ahead may require deciphering intricate patterns in data,” says Tark. “Making sense in that sort of uncertainty is very exciting.”