Nepali duo goes from Kathmandu Valley to Silicon Valley

Two young Nepalis have founded an AI company that is on the cusp of takeoff after getting funding from a top accelerator program in the United States.

Sudip Rokaya and Kartikesh Mishra did know each other in Nepal, but met as students at the Massachusetts Institute of Technology (MIT) in Boston. They are now co-founders at Lamina Labs in San Francisco, working on generative video.Their first product, Simi, generates whiteboard-style explainer videos based on a text prompt, and has received backing from YCombinator (YC), a top startup accelerator and quickly also secured further $3 million funding.

“In high school I did a lot of physics, and I was a visual learner so I really liked educational videos,” says Rokaya, currently on leave from his Math and Computer Science undergrad program at MIT. Rokaya was inspired by YouTuber 3Blue1Brown, known for his use of animation and visuals to explain Math topics.3Blue1Brown wrote Manim, a custom set of tools in the programming language Python that he uses to create animations. This set of tools, called a library, are available for anybody to use or modify.

Last summer, Rokaya was trying to see if he could automate this library to also create videos but he ran into problems because Manim was designed before the Large Language Model (LLM) era, for users experienced in Python.

“Trying to force it to only do this small library is difficult, it would take millions of dollars to fine tune it to get very accurate,” says Laminal co-founder and Chief Technical Officer Kartikesh Mishra, who studied Electrical Engineering and Computer Science at MIT, and completed a Master’s in Engineering before moving to San Francisco.

Besides, there were other startups already working on the problem that had received funding, so they abandoned that idea. Three months into the semester though, Rokaya was taking a Biology class that was “not one of my favourites”.“I looked for an AI tool that could help me learn through video. None of what I found had particularly good products, and they didn’t seem to be improving much either,” recalls Rokaya. The problem seemed to be technical and not financial.

At the time, Rokaya was exploring using generative AI and agents to ‘vibe-code’ -- which involves using LLMs to rapidly generate and modify code and software.

‘AI coding agents are the single greatest force multiplier available to founders right now. I’m not exaggerating. The leverage is unreal,’ wrote Rokaya to his 5,000+ LinkedIn audience two months ago.

Sensing an opportunity, Rokaya dove into research in late October, skipping classes and concluded that the process could be sped up 50x, while also generating a better end product. He showed it to Mishra, who had worked for two startups in San Francisco. He was impressed enough to join up.

“One reason I came to the US was to gather knowledge in ways that I did not have access to in Nepal,” Mishra told us. “MIT was about learning, but in San Francisco I wondered if I should work on something the world needs.”BEATING THE COMPETITION

Their product, Simi, is named for Rokaya’s home village of Simikot in the remote corner of northwestern Nepal. “I needed a cool name, and the investors dig it,” he says.Simi is free to test for anybody online, and in a test could generate a one-minute explainer video about a concept in economics. The video was structured, with examples, intro, and even tips on how to apply the rule to real life. A talented human animator would have taken many hours to do a similar video.

Users may put in as simple or as complex as a prompt they want to generate the video. For the end product, users can even choose between three aspect ratios, a male or a female voice, 80 different languages including Nepali, and lengths of 1, 2, 3, or 5 minutes.In the cutthroat world of AI and technology, where solutions get copied instantly, companies need ‘moats’ that elevate their products above competitors. Simi’s impressive moat is that it is faster, cheaper, and generates better videos all at the same time.

From the quality standpoint, Rokaya and Mishra followed an intense iterative process that involved making small tweaks to the product, solving what was not working, and evaluating the end product.Simi is a ‘deep tech’ product, one that runs as close as possible to the machine, as opposed to an ‘LLM wrapper’ -- a product that uses other models under the hood. With attention span and patience dwindling, speed of generation is a major benefit. It does all of this while being a relatively small model, cheap to run. Says Rokaya: “As we are both from Nepal, we wanted Nepalis to be able to afford the tool, and deploy it through cheap servers which cuts costs.”

He believes that it could help teachers in isolated areas of Nepal like where he comes from.

While Lamina Labs initially thought education was where its product would be most used, the two ran into initial issues selling. “The education business is slow, so we targetted small businesses instead,” says Rokaya who was confident it was a unique tool suitable for any field.

YC’s first post about Simi went viral, and so has every subsequent update. Twelve days after launch, the company had 4,000 users, 187 of them paying customers.“Small and home businesses in Kentucky, Florida,” says Mishra. “Pet daycares are using our tool to make their clients videos about dog vaccinations.” One Saudi Arabian was on Simi’s max plan and used up his quota in three days.

Lamina Labs made news because it was the first all-Nepali startup to get funded with $500,000 from YC which connects its startups to an extensive network of mentors and a community of other talent.“The main thing is the community. We can text anybody in the system, like the CEO of Airbnb. And they are very active,” says Mishra. “It has an acceptance rate of less than 1%. But if you get in, the resources are close to infinite. The mentors know what works, what doesn’t.”Both point to the community, rather than the money, as most important. Once every two weeks they have one personal meeting with a YC partner -- in their case, a CEO of a company that went public — and one group office hour ‘pod’ with 12 other companies where they set goals of revenue, number of customers for their company.

“By setting these goals publicly, it makes us accountable and urgent,” says Rokaya. “It makes people 100x more productive and effective than they normally are.”Most recently, YC was holding Demo Day 8 - 16 June, where founders present their product in front of a thousand potential investors. While each company expects to talk to more than 100 investors before getting funding, the first investor that Lamina Labs talked to was ready to invest all of the $3 million they were asking for.“They told us not to raise with anyone else, and just get to work. So we are pretty relaxed about the event,” says Rokaya, “we are actually having to decline other investors.”With the funding secured, Rokaya and Mishra plan to hire more engineers. Up to now, it has just been the two of them.

Rokaya has cautionary advice for Nepalis in this sector. “People say they want to do startups, it sounds cool. But most of the time they are not ready at all,” he says. “So many people are in computer science. There are maybe 1,000 people trying to be successful with the same idea.”Rokaya says people need to be “locked in” with 24/7 focus. The two used to get up at 10, walk four steps to their desks, and work nonstop until 2AM. Mishra admits that is not the healthiest lifestyle, but says the field is so competitive you have to be ahead of everyone else.

The shared Nepali background has been a major help. “A lot of startups break down because the founders can not get along over time. Both of us being from Nepal helps in talking to and understanding each other, and working,” says Rokaya.