Regional AI models present vast opportunity

Focus must be on generating, paraphrasing, editing and refining Telugu, Tamil, or any other Indian language text to sound like a native speaker’s work

By Kattamreddy Ananth Rupesh

It is no shocker that India is lagging behind in the global AI and IT race, given our risk-averse attitude towards innovation and our reluctance to think in terms of large-scale, disruptive business models. However, this does not mean we lack opportunities.

---

One of the most underexplored areas is the development of Large Language Models (LLMs) for regional languages. While English-based models are already performing exceptionally well for average users and excelling in code generation for computational tasks, the same cannot be said for Indian languages. Fine-tuning LLMs on regional literature and datasets presents a vast opportunity. The ability to generate, paraphrase, edit, and refine Telugu, Tamil, or any other Indian language text to sound like a native speaker’s work is an immediate and achievable goal.

Core AI infrastructure
However, beyond language, India’s AI sector needs to focus on building core AI infrastructure rather than relying on open-source-based models with limitations. Developing our own AI frameworks ensures we retain control over innovation, avoid excessive royalties, and stay ahead with cutting-edge advancements rather than playing catch-up.

The legal battle over dataset training and royalties is only beginning, and recent events, such as the controversy surrounding Suchir Balaji’s case, highlight the turbulence ahead. India is set to witness a slew of litigation tangled in this domain for sure in the near future when countries start using litigation as a means to restrict or promote businesses.

India’s lag in AI and IT is rooted in several factors. Fragmented AI development is a key issue; unlike China, where AI research is state-backed and heavily funded, India lacks a unified approach, and much of our top talent is lost to global firms. Regulatory bottlenecks and policy indecision further slow down large-scale AI ventures.

Another major challenge is our insufficient investment in computing infrastructure; AI development is compute-heavy, and India is yet to establish large-scale AI-dedicated data centres. Moreover, our IT industry has long been oriented toward services rather than product innovation. While we have excelled as an outsourcing hub, we have yet to produce AI-first global products that shape the industry.

Talent export factory
One of the harshest ironies is that despite the government pouring billions into IITs and training some of the brightest minds, we have essentially become the world’s finest talent export factory. We train them, polish them, and then — like a well-oiled conveyor belt — ship them off to Silicon Valley, where they build trillion-dollar empires for foreign companies while India gets a LinkedIn post about their “humble beginnings.” It’s almost as if we are running the most generous scholarship programme for global tech giants, with no returns except an occasional patriotic TED Talk.

That said, there is a clear path forward. If India invests in regional AI models, sovereign AI infrastructure and regulatory clarity, we can take advantage of the massive linguistic and computational potential at our disposal. AI should not just be another industry where we provide services — it should be a domain where India leads with original innovation.

(The author is Assistant Professor of Forensic Medicine and Toxicology, Andhra Medical College, Maharanipeta, Visakhapatnam)