Perplexity AI CEO Calls for India’s Own AI Revolution

Perplexity AI CEO Aravind Srinivas believes India should develop its own AI models instead of depending on foreign technology. Speaking on Nikhil Kamath’s WTF podcast, he highlighted India’s untapped AI potential and suggested the country establish research firms similar to DeepSeek. These companies should focus on both Indian languages and global AI benchmarks, inspiring young engineers to innovate. Srinivas outlined a roadmap for AI startups, advising them to secure funding, grow their user base, and invest in infrastructure.

He recommended starting with post-training on open-source models before moving to pre-training. A key gap he identified is voice recognition and synthesis for Indian languages, as current AI models struggle with Indian accents and dialects. Since Western AI research does not prioritize Indian linguistic diversity, real-time AI voice technology tailored for India could be a game-changer. Despite challenges in collecting and training AI voice data, he sees this as an opportunity for India to gain a global competitive edge in AI.

Perplexity AI CEO Calls for India’s Own AI Revolution

Aravind Srinivas, CEO of Perplexity AI, has urged India to prioritize building its own artificial intelligence (AI) technologies instead of relying on foreign-developed tools. Speaking on Nikhil Kamath’s WTF podcast, he emphasized India’s immense yet largely untapped potential in AI. By fostering local innovation, he believes the country can address its unique challenges while also competing on a global scale.

Why India Needs Its Own AI Models

Srinivas argued that India’s diverse linguistic and cultural landscape requires AI solutions tailored to its needs. While global AI models like ChatGPT or Gemini are highly advanced, they often struggle with India’s 22 officially recognized languages, hundreds of dialects, and regional accents. For example, voice assistants like Siri and Alexa frequently misinterpret Indian English accents and regional languages such as Tamil or Bengali. This gap exists because Western companies primarily design AI for their own markets, leaving India’s needs underserved.

To address this, he proposed establishing Indian AI research firms modeled after companies like China’s DeepSeek. These homegrown startups could develop models that excel not only in Indian languages but also on global performance benchmarks. By creating AI that understands local contexts—from agriculture to healthcare—India could inspire young engineers to solve the country’s unique problems while contributing to global AI advancements.

A Roadmap for Indian AI Startups

For Indian entrepreneurs entering the AI space, Srinivas outlined a clear strategy. First, secure funding to scale operations, as investors are increasingly interested in AI. Next, focus on expanding the user base, as a larger audience provides valuable feedback for refining products. Finally, reinvest profits into research and development to ensure long-term growth.

He also shared a technical blueprint:

1. Start with Post-Training – Customize existing open-source AI models (such as Meta’s Llama) for specific tasks. For instance, an AI language model could be fine-tuned to answer questions about Indian law or education.
2. Move to Pre-Training – Once expertise is gained, develop original AI models from scratch. This involves training AI on large datasets—including Indian textbooks, news articles, and regional language content—to build a deeper understanding of local nuances.
3. Invest in Infrastructure – Build data centers and acquire high-performance computing resources, which are essential for efficiently training advanced AI systems.

The Voice Technology Opportunity

Srinivas identified voice recognition and synthesis as a critical area where India lags behind. Many AI tools today struggle with Indian names, accents, and dialects. For example, a farmer in rural Uttar Pradesh might find it difficult to use a voice-based AI app designed for American English speakers. Addressing this challenge, he said, could transform industries like healthcare, education, and customer service. Imagine an AI tutor that teaches math fluently in Marathi or a telemedicine app that understands rural dialects.

However, collecting voice data from India’s diverse population is complex. It requires recording thousands of hours of speech across different regions, ages, and dialects. Privacy concerns and technical challenges—such as filtering background noise in crowded areas—add further difficulty. Yet, Srinivas views this as a golden opportunity. Startups that develop real-time, multilingual voice AI could not only dominate the Indian market but also export their solutions to other linguistically diverse regions like Africa and Southeast Asia.

Challenges and Global Ambitions

While the vision is promising, challenges remain. Funding for AI startups in India still lags behind Silicon Valley, and top talent often migrates to global tech hubs. Additionally, training AI models requires significant computational power, which is expensive and energy-intensive.

Despite these hurdles, Srinivas remains optimistic. He believes India’s scale and diversity are advantages. By leveraging its vast population, startups can collect unique datasets to train robust AI systems. For example, an app designed to transcribe Indian languages could improve rapidly with input from millions of users across the country.

Conclusion

Aravind Srinivas’s message is clear: India’s AI journey must be self-driven. By focusing on local challenges—such as language barriers—and investing in homegrown innovation, the country can establish itself as a key player in the global AI landscape. Locally developed AI models could empower millions of Indians by providing technology in their native languages, enhancing digital inclusion. Moreover, success in AI could inspire a new generation of engineers, much like India’s IT boom did decades ago.

The road ahead is challenging, but the rewards are immense. As Srinivas put it, “India isn’t just a market for AI—it’s a laboratory for the world.” By embracing this role, the country could transition from a consumer of foreign technology to a pioneer of inclusive, locally rooted AI solutions