Kathmandu Valley to Silicon Valley

Generative AI models such as ChatGPT and Sonnet are trained on massive amounts of data that have over time enabled them to answer general questions. But they do not perform as well on narrower domains, even hallucinating or making things up.

Now here comes Anthromind, a startup developed by a 28-year-old Nepali engineer Pratik Karki in the tech-womb at Silicon Valley, which develops datasets to improve the performance of AI models in specific areas.

Large language models are ‘distilled’ by running them on smaller, higher-quality data. AI is first used to assist with labelling or generating data. After that, human experts look at the responses and explain how they are wrong, helping tune the model just right.

Karki got the idea while working at Google, where he saw that while companies were trying to create their own models, they failed because the data was of low quality. In 2024, he left his lucrative job and took a risk by setting up a business that could make real impact.

It has not been easy. At Google he was one of over 100,000 employees, but after launching Anthromind, Karki has had to do sales, marketing, and recruiting, as well as leading a whole new team.

“At some point in your career, everything turns into sales,” Karki told us by phone. “You need to sell potential investors on your vision, and you get a million no’s before a yes.” The connections he made at Google helped, and he even managed to poach a few and others became investors.

Karki studied in Biratnagar before moving to Kathmandu to do his A-levels where he did business courses. He remembers those formative years and credits his teachers in Nepal for encouraging him to put himself out there and to speak up when needed.

He then went to Grinnell College in Iowa where he studied computer science and statistics, and realised that he liked to code: “I especially got a high debugging a program until it worked.” 

He went on to pursue a Master’s in Computer Science and describes the course as a pure program with some very difficult machine learning courses.

Read also: Tech and transition in Nepal, Ashutosh Mani Dixit

Karki was recruited to Google from a highly selective early career program and got to work closely with the companies that were then building the generative Large Language Models that have since taken the world by storm. He worked on essential training infrastructure, including data warehousing and computing needs.

“Internally, we all knew about this technology and how disruptive they would be,” says Karki. “But I also saw its limitations.”

Karki took a calculated risk in setting up his own company, knowing fully well that for Anthromind to succeed he would need a very obsessive and disciplined mindset. 

“Every Friday evening I look at weekly data, and what worked and what didn’t, especially in terms of getting customers,” he says. “I had to be very deliberate with everything I did.”

Karki attributes this discipline and frugality to growing up in Nepal, which he misses dearly, especially his family and the community. His best friends in California are Nepalis, they meet often and have already made plans to celebrate the approaching Nepali New Year with home-made momos. 

Karki sees a lot more potential in the IT industry in Nepal, at the moment saturated with service-based companies when ideally new businesses should focus on products. 

He explains how companies that provide offshore tech services operate in a ‘default:alive’ state, where they can carry on even if they only have one client.

The alternative is a company whose success depends entirely on the quality of the product it puts out, a ‘default:dead’ state. This forces programmers to adopt an urgent, disciplined mindset. While a lot harder and riskier, it also has higher returns.

Read also: Nepal as an AI power bank, Bikash Pandey

Karki is in a cohort of young Nepalis in the IT industry who have proven that serious tech innovation can come out of Nepal. Pukar Hamal of SecurityPal was recently featured in Forbes, while Sameer Maskey’s FuseMachines has been listed on NASDAQ.

Financially independent since 18, Karki says Nepali families should encourage their children to make such moves. Nepal’s education system should reward curiosity, instead of quelling it, to promote entrepreneurship. 

He says, “Students should be rewarded for thinking individually and speaking out, like I was.”