3 days ago
Historian Sampath, technologist Chauhan build AI tool for faster book translations
Representative Image
BENGALURU: Bengaluru-based Naav AI, an artificial intelligence startup co-founded by historian
Vikram Sampath
and technologist Sandeep Singh Chauhan, has spent the past few months in stealth.
But the problem they're chasing is hiding in plain sight: India has too much English and too little access.
'A book like Savarkar's took nearly two years to appear in Marathi,' Sampath, whose experience with delayed translations across languages like Hindi, Marathi, and Kannada directly shaped Naav's mission. "Only 5-6% of India reads in English. Yet English dominates everything, books, media, even AI training data."
Naav's first product, TransLit, targets this imbalance. It blends multiple large language models (LLMs) and a proprietary workflow to rapidly translate long-form text across six Indian languages – Hindi, Marathi, Kannada, Tamil, Telugu and Malayalam. An average 300-page book, Sampath claims, can now be translated to first draft in under an hour. Human editors then step in through a custom dashboard, refining output line-by-line and feeding their changes back into the engine.
Chauhan, a former senior tech executive who led digital transformation at the Technicolor Group, said Naav isn't trying to build a foundational LLM. Instead, the team is building agentic workflows that orchestrate multiple public and private models depending on language and context. "We've seen models like Claude work well for Hindi. But for southern languages, accuracy drops to around 50%. That's where our orchestration and feedback loops come in," he said.
The system currently achieves around 60-65% base accuracy, with ambitions to push toward 80%.
The founders are quick to clarify that Naav is not about replacing human translators. "This is not a zero-touch translation," said Sampath. "The goal is speed and scale, not displacement."
Naav AI has raised early backing from Bhavish Aggarwal and Silicon Valley investor Asha Jadeja Motwani. Its initial client is publishing house BlueOne Ink, which has committed a pipeline of 30 books.
Of these, 18 are already in production.
Beyond text, Naav is eyeing audio. Its second product, ZuNaav FM, is being built to generate immersive, multilingual audiobooks and thematic content using voice synthesis and background engineering. "Imagine listening to Tipu Sultan's biography narrated in my voice, with war scenes playing in the background," Sampath said.
The startup's next step is scale. For now, Naav runs a service model with in-house and contract language experts. Eventually, the plan is to offer it as a software-as-a-service platform to publishers and enterprises.
"We're not just translating text," Chauhan said. "We're translating access."
Stay informed with the latest
business
news, updates on
bank holidays
and
public holidays
.
AI Masterclass for Students. Upskill Young Ones Today!– Join Now
