Kerala Technology
Kerala student embodies Indian AI talent

Govind SB speaks during a recent MuLearn event at Techno Park. Photo courtesy: MuLearn

Kerala student embodies Indian AI talent

Hari Kumar By Hari Kumar, on June 25, 2024
Hari Kumar By Hari Kumar, on June 25, 2024

On a day of torrential rain, when a youngster walks into a cafe for a meeting fully drenched, with water dripping from his shirt and messy hair, you instantly realise two things: this guy doesn’t care about what others think, and he is determined.

While these traits have led many youngsters down paths that end in swampland, Govind SB has chosen a direction that looks to discover new horizons, like artificial intelligence.

The final-year computer science student at Marion Engineering College in Trivandrum has already gained some attention by winning the top rank at the Top 100 Coders 2023 event held by Kerala Startup Mission in association with MuLearn to identify the best coding talents in the state.

Now, the 21-year-old is being consulted by companies looking to integrate AI into their businesses, and he has launched a consultancy firm called AIC Project to run such projects.

At a time when there is a worldwide hunt for AI expertise, Govind exemplifies the kind of talent India can offer. His skills represent the potential among Kerala’s youth and suggest a promising future for the state in the global IT landscape.

Govind dismisses the courtesy tea/coffee offer with an “I-am-a-no-coffee-tea-person” answer as he sits down. His scraggy beard indicates a rebel, but this guy is a rebel’s rebel. During a recent event at Techno Park, he said he shuns online entertainment like Netflix, and urged the mostly-student audience to go home and watch YouTube videos of experts like Canadian computer scientist Andrej Karpathy.

As the conversation starts, his staccato explanations turn into lengthy rat-a-tats when they delve into computers and technology. Despite his bursts of nervous energy, Govind says he is very patient when he approaches projects, especially those that involve emerging technologies.

He attributes his patience to the slow internet that was around during his school days. He spent three months downloading GTA-5, a game that was 60GB, using BSNL internet lines with a download speed of less than 10 Mbps. “But I learned about download managers and how to pause and continue downloads later,” he says as an afterthought.

Govind says he was a computer child from an early age, having access to a machine even before he was ten years old because his father, Sajith Kumar, was a lecturer at the Institute of Human Resources Development colleges and had a computer at home.

“Those days, BSNL had an offer that allowed unlimited usage between midnight and 6am. Every night, I would wake up at midnight, walk to the computer like a zombie, and start surfing. I think I was the only person who made use of the BSNL offer every night until they pulled the plug on it years later,” says Govind, squinting through his foggy glasses as if trying to locate a bit from a hazy past.

As with any kid of that age, his main interest was in games. While most took pleasure in playing them, Govind says his challenge was figuring out how to download them or crack the codes. “Sometimes, my interest ended as soon as I downloaded the game.”

Around the time he was in the 11th grade, blockchain technology took off, and Govind jumped headlong into it. That naturally led him to the crypto world.

Initially, he borrowed some money from his dad and started trading, promptly losing all of it. Then a Brazilian friend he met over the internet gave him 50 US dollars to trade and learn about the crypto scene.

Soon, Govind developed an algorithm to trade crypto and even managed to get some clients for his investment project.

“At one time, I was managing 100,000 US dollars for my clients, all while I was 14 years old. Since it was crypto, everyone was anonymous, and no one knew my age. They only knew my profile picture on Discord, and they didn’t care as most of them were making profits of around 600 percent.”

But when the Indian government started bringing in restrictions on crypto trading, Govind handed over the investment platform to the members and quit the scene.

“I must have made around 3.5 lakh rupees in that trade, and I bought a powerful machine that was needed to do crypto. It was worth a lot then, but now with the crypto crash the price of such models has come down to around 60,000 rupees.”

Along with crypto, he also helped people who wanted to develop products and operating systems using blockchain technology.

His exploration of the gaming world continued, and he stumbled onto a game called AI Dungeon. As he learned more about the game, he became curious about its inner workings, which led him to the discovery of GPT 2 and 3.

“There was no one here who could tell me about it as very few had even heard about it.”

When Stable Diffusion released its picture generation tool, Govind became fascinated with it and absorbed all the chatter about it on the internet. The teenager in him led him to find ways to make Stable Diffusion create images that the tool was generally restrained from.

There were very few from Kerala or India posting tips on this, and Govind says he had to scrape bits from forums like Reddit and 4Chan to get what he wanted. After creating a few images, his next quest was to create stories and then generate images to go with them.

“But I was too lazy to write scenarios, so I looked for ways to automate it. That's when I started learning about LLMs. So, my journey into AI models was kind of the opposite – from image generation models to the discovery of LLMs.”

Then ChatGPT burst onto the scene, followed by models like Lama2, and soon there were several LLM models, small and big, available. While were majority were experimenting about writing prompts and creating images, Govind’s aim was to try and figure out how to run these LLMs in his machines to learn about their structure and inner workings.

Govind says he had a network of local IT nerds developed through hackathons and similar tech events, but those who were into LLMs were limited.

By the time he entered his third year, he started going to Tinkerspace in Kochi, the only hacker house in Kerala that was open to techies around the clock, seven days a week. It had some cutting-edge computers, including machines with high-end graphics cards, headsets, 3D printers, and GPUs essential for running AI experiments.

“At that time, I was the only one using the GPU to run AI models, and initially, I was a novelty. As I was constantly updating myself on the latest happenings in the AI field, that made me a popular figure there.”

“I used to walk from the Metro station to Tinkerspace in Kalamassery to save on the auto fare of 100 rupees. I was going there every weekend, so that would have come to around 1,000 rupees for autos. Too much for a student like me.”

As his journey into AI continued, in 2023 Kerala Startup Mission in association with MuLearn decided to hold a contest to find the top 100 coding talents in the state. After hearing about his exploits, Deepu S Nath, who manages MuLearn, decided to give direct entry to Govind.

The Top 100 Coders event had attracted talents of all ages from across the state, and Govind, true to his traits, decided not to try his hand at AI. Instead, he chose to take a shot at Beckn Protocol, the open-source platform for peer-to-peer networks.

“In my mind, I wasn’t there to win the prize. So when they offered me a chance to try my hand at something new, I picked that,” says Govind.

After the contest that lasted for weeks, Govind secured the top rank as the best coder in the state. “The Beckn code was huge, and we didn’t have machines to run that kind of program. So during the final interview, I told the judges straightaway that the code would not work, but I explained how the code should work and why it works that way.”

Govind says he loves analysing problems and solving them. At his college, he discovered that seating candidates in the exam hall was a time-consuming process for the faculty, as they had to manually create a plan to avoid grouping together students taking the same paper.

“I learned that my teacher was spending two weeks to come up with the plan. So I worked out a plan and developed a program for it. All she had to do was enter the data, and it would be done in five minutes.”

“But I’ve yet to get any brownie points for that,” he says ruefully. “I had to struggle even to have my attendance sheet validated.”

Meanwhile, his interaction with MuLearn increased, bringing together like-minded friends and opportunities such as hackathons organised by the Kerala Product Hunt group. A stint as an LLM Development Intern at the Tinker Lab also bolstered his online profile and added substance to his CV.

He continued posting about his experiments with AI and LLMs platforms like Github and Twitter (now called X), effectively serving as a reference guide for companies seeking expertise to solve their problems.

A meeting at the Product Hunt hackathon with Jijo Sunny, the founder of Buy Me a Coffee, later led to Govind's first official gig as an AI consultant. He spent a few days at the Buy Me A Coffee backend office in Kochi and worked with them to develop the IT architecture for their new application called VoiceNotes.

Following this, he received an offer from Hoppscotch, a London-based open-source API developer, as a product engineer, which pays him enough to take care of his needs as a student.

A couple of other gigs are in the pipeline, and Govind is waiting for the formalisation of agreements before he can make them public, including one with a unicorn company familiar to the Kerala tech community.

Hours pass by as Govind talks about his ideas for products like an AI companion that you can hang around your neck, an app to remind you to call long-lost friends, his theory of how brain implants could be the first step in developing androids, how AI is being hyped too much, and how he is buying Nvidia shares to secure his future.

He also discusses his concerns about job losses and advises youngsters on how to prepare for the future. Now that he is a “veteran” AI guy, what advice can the 21-year-old give to his peers?

“If you are not yet in the AI field, get into it. There are a lot of opportunities. There are a lot of problems that need to be solved in AI. Data preparation, data cleaning up, synthetic data generation, how to inspect data… Nobody knows about this.”

“Age doesn't matter here; I am just 21 years old. Anyone could have done it. Why no one did, I have no idea.” Don’t be afraid to ask dumb questions and don’t be afraid to test out your wild ideas are his guiding mottos.

He says he knows students in Kerala who started learning about AI after the release of ChatGPT and have emerged as top talents in the field. But he admits there needs to be laser focus and determination.

Obviously, those are the qualities that Govind displayed when he walked into the cafe despite the heavy downpour that may have deterred many others.

 


 

Alcohol, cigarettes and … social media?

Should social media sites carry a mandatory warning? Yes, says the US Surgeon General, Vivek Murthy, who urges a mandatory warning for social media similar to the labels you see on alcohol and cigarettes. Most of these concerns about internet addiction stem from rising mental health issues among young people. A recent study in the Indian Journal of Psychological Medicine was different as it looked at internet addiction among adults in the country. It found that 58 percent were addicted, with 0.88 percent severely addicted. While authorities, especially in Kerala, are quick to blame the youth for societal problems, problems stemming from the adult population are hardly tackled. As the study suggests, more research on internet addiction among adults in India is needed.

Restricting the use of social media is a tough challenge, and one US-based startup, which lists IIT Bombay alumnus Vikram Chauhan among its cofounders, offers a way to reduce your scrolling time. The Roots app provides an analysis of your scrolling time and grades it based on the quality of your interactions. The app makers claim that identifying “digital dopamine” will help you avoid wasting time on your mobile phone with mindless scrolling. It “lets users lock themselves out of the most addictive apps and adds reminders to stop obsessively scrolling through algorithmic social feeds”,  reports TechCrunch.

 


 

Indian gaming company eyes SE Asia

Indian e-sports continue to grow in strength, and local companies are now eyeing opportunities abroad. Ahmedabad-based company Gamerji says it is looking towards Southeast Asia to expand its audience, beginning with Indonesia and the Philippines. Their platform currently offers eight game titles, including PUBG Mobile, Free Fire, MLBB, Valorant, Clash Royale, CS, Rocket League, and FIFA. The company plans to add one million users from the Southeast Asia region over the next 12 months, reports the Financial Express.

 


 

Gemini says Namaste to India

Google has released its AI chatbot, Gemini, in India, which supports nine Indian languages, including Malayalam and Tamil. The company says people can chat or speak in those languages to use Gemini. According to the company's blog, people can use the chatbot to organise a get-together, find a recipe, or even upload a picture of a flat tyre to get help with the steps to replace it. Our initial cursory glance showed that it will need to overcome some teething problems, but that is expected. If Gemini proves to be effective, it will bring millions of mobile phone users in India face-to-face with AI quickly.

 


 

A purrfect startup for Japan

We all know Japan is an ageing society, but the latest alert from the Pet Food Association in that country is a bit worrisome. It says there are 16 million pet cats and dogs in Japan, which is more than the number of children under 15 in the country. A startup there has tapped into this and come up with CatsMe, an app that helps owners know if their cat is well or not. With the help of researchers at Nihon University, the firm created the app by training it on 6,000 pictures of cats to analyse their behaviour and symptoms. The app has been embraced by 230,000 customers since its launch last year, reports Reuters.