IMAGE: TINGTING
IN EARLY February, at a conference centre in Doha, a young Nepali engineer named Anjelika Sah walked onto a stage and presented a piece of software built entirely in Nepal to an audience of over 700 international investors and delegates from 124 countries. The event was Web Summit Qatar 2026, one of the largest technology gatherings in the world, with more than 1,600 startups competing for attention. Her company, TingTing, was the only startup representing Nepal. It had been incorporated for less than a year, though the core technology behind it had been in development for longer.
TingTing is, in the driest possible description, an AI-powered business communication system: a piece of infrastructure that sits between a company and its customers, handling voice calls, messages, emails and text interactions through a single unified dashboard.
What makes it interesting, and what got it to Doha, is the technology underneath. The company has built two proprietary systems from scratch in Nepal: the RIRI engine and the Understanding Engine. RIRI is, at its core, a text-to-speech system: it converts written Nepali into natural-sounding spoken Nepali, capturing local tone, emotion and linguistic rhythm in a way that existing global platforms have never bothered to achieve. The engine has evolved through three versions: Riri-1, which turns text into speech; Riri-2, which converts speech back into text; and Riri-3, a generative AI model capable of holding a two-way conversation with a customer without a human on the other end. The Understanding Engine sits alongside, interpreting the intent behind what a customer says rather than just recording the words. Both were built and trained in Nepal, on Nepali data, by a Nepali team.
To understand why this matters, consider what AI voice systems typically look like when they encounter a language like Nepali. The major global platforms such as Google and Microsoft have built voice engines trained overwhelmingly on English and a handful of high-resource languages. When a Nepali bank tries to use one of these platforms to deliver automated calls to its customers, it gets a voice that technically speaks some Nepali but sounds foreign and emotionally flat. Customers hang up. The product does not work. TingTing’s pitch is that Riri sounds like a person: the company trained it on multiple voice-over artists, worked on emotional nuance and built it specifically to carry the natural rhythm of spoken Nepali. Traditional voice systems check roughly 2% of calls manually; TingTing Connect analyses every one in real time, tracking sentiment as the conversation moves.
The domestic record supports a claim of genuine capability. Before going to Doha, TingTing had signed with WorldLink, Nepal’s largest internet service provider, Citizens Bank, eSewa (the country’s leading digital wallet) and Suryodaya Insurance. It handled the AI-powered information calls for the 1974 AD concert, a large Nepali music event, in what the company described as Nepal’s first AI-powered mass event communication campaign. It signed a contract with Dulikhel Municipality for public-sector communication. In March it launched what it called Nepal’s first AI-native call centre solution—offered initially at no cost to enterprise clients—which automatically records every call, transcribes conversations in real time and analyses customer sentiment as the interaction progresses. The launch also introduced “The Box”, a proprietary hardware unit designed to keep all communication data within a company’s own network rather than sending it to an external cloud.
The international expansion has already begun. In May TingTing signed a partnership with Effective Solutions, a Sri Lankan company, to deploy its system for businesses and public institutions there. An office in Colombo is in progress. The expansion reflects an intentional strategy: entering markets that share the linguistic complexity that made Nepal a useful testing ground, and where the major global AI platforms are similarly absent. If the RIRI engine can be extended beyond Nepali—the company demonstrated at Web Summit that multi-language expansion is underway—TingTing has a template that could work across South and Southeast Asia, where dozens of languages of tens of millions of speakers exist in a similar state of neglect by the large AI companies.
There are two real risks in this strategy. The first is that the global AI giants could invest seriously in low-resource language voice engines if the market became large enough to justify it. The counter to this risk is that TingTing is not trying to beat Google at a global language: it is trying to be the trusted local expert that a global giant will never bother to become. That is a viable position, but it depends on moving fast enough across enough languages to build a defensible position before anyone else notices the same opportunity.
The second risk is more immediately practical. TingTing is operating on an introductory offer: services at low or no cost while building its client base, with a paid model to follow. This is a reasonable market-entry approach for an enterprise software company that needs to prove its reliability before asking for a contract. It is also an approach that burns cash, and the company has not disclosed its funding position or runway.
It arrived in Doha without a publicly announced investment round. At a conference with 700 investors in the room, the conversations that followed the Startup Showcase were, by the company’s account, encouraging. What converts from encouraging to a term sheet is what the next several months will reveal.
What TingTing’s Web Summit moment tells you about Nepal’s technology sector is something more than one company’s marketing trip. It is evidence of what the country’s engineering graduates—the 10,000 who leave university a year with computer science and technology degrees—are capable of building when they stay. The problem is that most of them do not stay. They take jobs in Australian technology companies, American consulting firms or Gulf-based IT departments, sending money home in the same remittance pipeline that funds everything else in Nepal. TingTing’s team built a production-grade AI voice engine in a country where the standard assumption is that serious AI work happens somewhere else.
The country needs its own version of the story TingTing is trying to tell: that a technical problem specific to this part of the world can be solved here, by people who grew up here, building products that the world will eventually pay for. The alternative—that every Nepali engineer with the skills to build something like RIRI leaves for a city that already has ten companies building the same thing—is the version of events that has been playing out for twenty years.
Anjelika Sah got on a stage in Doha and told the world that version was changeable. Proving it will take considerably longer than a four-day conference. ■
