Vikram Sampath’s NAAV AI sets sail to bridge India’s language gap with human-centric AI

NAAV AI founders Vikram Sampath (left) and Sandeep Singh.

At a time when generative artificial intelligence (Gen AI) is reimagining everything, from customer support to movie scripts, a Bengaluru-based startup is applying it to a different frontier — India’s linguistic soul.

Co-founded by historian-author Vikram Sampath and technology executive Sandeep Singh, NAAV AI aims to make high-quality translations and audiobooks in Indian languages not just possible but scalable, precise, and deeply cultural. And it’s doing so with a mix of AI and human insight.

The company formally launched on June 5 after operating in stealth mode for nearly five months. Incorporated in January, NAAV AI, which is short for "Navigating AI Across Vocabularies" and means boat in Hindi,  aims to solve a long-standing problem in India’s publishing and content ecosystem — the massive gap between the demand and supply of high-quality content in Indian languages.

“Most of my books, to get those out into Indian languages, it was taking such a long time, sometimes years,” said Sampath, who has authored biographies of Gauhar Jaan, Vinayak Damodar Savarkar and Mysore Kings. “There is demand, but the supply chain is broken,” Sampath told Moneycontrol.

India’s publishing market, valued at $10.7 billion in 2024, is expected to grow significantly, fuelled by regional language content, rising literacy, and digitisation, according to an Ernst & Young report. With 9,000 publishers and up to a million books published annually, the country ranks second globally in English publishing. Yet only a fraction of this content gets translated into Indian languages, said Sampath.

This is where NAAV AI comes in, with two products — TransLit and ZuNAAV-FM.

TransLit is an AI-assisted, human-refined translation tool that currently supports English-to-Hindi, Marathi, Kannada, Tamil, Telugu, and Malayalam translations. Unlike Google Translate or DeepL, TransLit is built to handle large documents and books, offering contextual accuracy and editable, well-formatted output.

“What makes TransLit different is how it handles entire books page by page while retaining context,” Singh said. “You upload a manuscript, get an initial AI translation, and then a human expert refines it.”

TransLit is built on agentic frameworks and a mix of existing Indic LLMs, such as those developed by IIT Madras, combined with custom logic for weighting and selecting outputs.

“We use a mix of models and weigh their outputs to arrive at the best translation,” Sravan Kumar Aditya, senior technology adviser at NAAV, told Moneycontrol. “Even existing LLMs struggle with context beyond a few paragraphs. So, we’re also building our own corpora with licensed data from books and professional linguists.”

Sampath said the results are already showing. “Initially, Marathi translations had just 45-50 percent accuracy,” Sampath said. “Now, with feedback loops and linguistic tagging, we’re seeing major improvement.”

Addressing concerns around ethics and bias in AI systems, the core team at NAAV stressed its commitment to transparency and quality. “We are building a licensed corpus. We’re not scraping the web and every output is reviewed by experts,” Aditya said. "Hallucinations are common in LLMs when they lack context. That’s why our feedback loops and human editors are so crucial," he added.

On the ethical front, Sampath said, "There’s a fear that AI will replace creative jobs but we believe it will only make people more efficient. No machine can replace human creativity."

NAAV AI has completed translations of three children’s books on Indian history into six languages (18 books in total) in collaboration with a private, non-profit Foundation for Indian Historical and Cultural Research (FIHCR) and publishing partner BluOne Ink. The books will launch in July.

NAAV AI operates on a service model, charging per-word translation rates that are lower than traditional human workflows. But the team sees future potential in SaaS and subscription-based models once publishers gain confidence in the platform’s output.

The idea for NAAV AI was born when Sampath struggled to get his 900-page book on Tipu Sultan translated into Indian languages. “By the time translations are done, the buzz around a new book has died down,” Sampath said. “It’s a problem for authors and publishers alike.”

That realisation led to brainstorming with friends and now colleagues, Singh and Aditya. Together, they saw an opportunity to bring together cutting-edge tech with India’s multilingual cultural richness.

Beyond books, the team sees use cases across media such as real-time translation for government services and education in regional languages. “The NEP (National Education Policy) mandates education in mother tongues,” said Sampath. “We want to support that.”

ZuNAAV-FM, NAAV’s second product, is focused on audio content. It aims to become an AI-powered audio platform for audiobooks and podcasts, all in Indian languages.

“We don’t just want to convert text to speech. We want to make audio immersive,” Singh said. “So, we are using AI for voice synthesis but also involving audio engineers to refine outputs, add effects, and preserve cultural nuances.”

The product is still under development and NAAV is looking to raise $1 million in funding to accelerate ZuNAAV’s R&D. The company has already secured $200,000–$250,000 in early-stage funding from Ola founder Bhavish Aggarwal and Silicon Valley-based venture capitalist Asha Jadeja Motwani.

A key philosophical pillar of NAAV AI is its commitment to "human in the loop". “You can’t automate creativity,” said Sampath. “We use AI to reduce turnaround time but professional translators are indispensable.”

"Our linguists tag and correct every output, feeding back into the model. The idea is not to eliminate human creativity but to enhance efficiency. A translation that took a month now takes under 10 days," Singh said.

The company’s core team includes Sampath, Singh and Aditya, supported by language experts and software developers. “We have one language expert for each supported language, with plans to scale as demand grows,” Singh added.

As India debates language politics, NAAV finds itself at an intersection of cultural identity and AI. Based in Karnataka, a state that has seen tensions over linguistic pride, the founders believe NAAV can be a bridge.

"Language should connect, not divide," said Sampath. "Our states were created on linguistic lines, but we should be able to interface with each other through technology."

Sampath also weighed in on Union Minister Piyush Goyal’s recent remarks on the quality of Indian startups. "We are at a stage where everything is required. Delivery startups, tech R&D, and content startups each solve a different problem. Startups also generate jobs. It’s not fair to pit one against the other."

Over the next five years, the team sees TransLit becoming the go-to tool for Indian and international publishers, with plans to expand into Arabic, French, and Spanish.

“The world wants to hear Indian stories in Indian voices,” said Sampath. “This is also about soft power.”