Synopsis
Articul8, a spin-off from Intel, is raising $30-50 million to establish an India-based AI entity focused on traditional knowledge systems like Sanskrit. The venture aims to build "heritage models" trained on ancient Indian languages and cultural data, with potential applications in various fields and global preservation efforts.Listen to this article in summarized format
Articul8 wants to build "heritage models" trained on ancient Indian knowledge systems through this venture, for which it is in initial talks with investors including Peak XV Partners, Accel, family offices linked to large industrial groups, and strategic investors such as Aditya Birla Group, which is already a backer of Articul8, the people said.
The proposed venture will develop multimodal AI models trained on Sanskrit and other classical Indian languages, combining text, voice and image capabilities. Heritage foundation models are specialised AI architectures trained on curated datasets aimed at cultural and historical preservation.
Articul8 has signed an exclusive partnership with Madras Sanskrit College as part of the initiative and plans to involve scholars, linguists and domain experts directly in model development, founder and CEO Arun Subramaniam told ET. He declined to comment on the fundraising discussions. Prior to founding Articul8, Subramaniam led Intel's Cloud and AI Strategy team.
The initiative comes as India's AI ecosystem increasingly shifts its focus from sovereign large language models to domain-specific and culturally rooted AI systems. Several institutions, including IITs and research organisations, have begun digitising and annotating historical texts, while startups are experimenting with Indic-language AI models.
Articul8's proposed structure would allow intellectual property generated through the heritage initiative to remain in India while licensing core technology from the parent company, Subramaniam said. The company plans to commercialise the technology through partnerships with businesses that already operate in consumer markets.
Queries sent to Peak XV and Accel remained unanswered at press time.
Global demand for heritage models
"The idea is to build foundational heritage models that can deeply understand grammar, semantics, context and oral traditions. Once we demonstrate that this architecture works for Sanskrit, the same framework can potentially be extended to other historical and civilisational languages such as Hebrew, Arabic, Parsi or even classical East Asian languages," Subramaniam said.
He said the company is collecting large volumes of Sanskrit data and encoding Panini's grammar rules into the architecture. The models will be trained on manuscripts, commentaries, spoken recitations and structured academic datasets.
Subramaniam said interest in heritage AI models is rising globally as countries and institutions look to preserve ancient knowledge systems and unlock historical archives that conventional AI systems struggle to process.
"There is a global recognition that modern AI systems are heavily trained on contemporary internet data and do not adequately capture historical or cultural intelligence," he said. "Many countries are now thinking about how to preserve their linguistic and cultural heritage using AI, whether it is around ancient manuscripts, indigenous languages or oral traditions."
Beyond language preservation, the models are being designed for applications in drug discovery, metallurgy, chemistry, healthcare, education and cultural preservation. The company is also exploring partnerships with government stakeholders for cultural and diplomatic use cases.
The heritage initiative comes as Articul8 expands its core enterprise AI business globally. The Santa Clara-based company raised the first tranche of a $70 million funding round in January 2026.
Articul8 develops specialised AI systems for sectors including energy, manufacturing, aerospace, semiconductors and financial services, enabling customers to deploy AI applications within their own IT environments. The company says it has crossed $100 million in total contract value and counts Hitachi Energy, AWS, Franklin Templeton and Intel among its customers.
