Krutrim, the AI startup founded by Bhavish Aggarwal, the CEO of Ola Group, has launched India’s first multilingual large language model (LLM) capable of generating text in 10 Indian languages. This voice-enabled LLM, known as Krutrim LLM, can understand various languages as well as a combination of languages known as Hinglish, which is a mix of Hindi and English. The startup is aiming to develop data centers and create servers and supercomputers for the AI ecosystem. Krutrim plans to release a chatbot in beta version in January, with prototypes of the servers expected in mid-2024 and production set to begin by the end of 2025.
The development of large language models in Indian languages, known as Indic LLMs, is garnering attention among startups and academic groups in India. These initiatives are driven by the desire to build their own AI systems rather than relying on technology from the United States or China. France’s Mistral AI, valued at $2 billion, and the United Arab Emirates’ Falcon model are some of the other models gaining traction in this space.
India, with its vast population of 1.4 billion people, is focused on building smaller and more cost-efficient AI systems. Sarvam, a generative AI startup, recently launched OpenHathi, an open-source Hindi LLM, after securing a $41 million investment. This announcement closely followed the completion of a funding round involving Lightspeed Venture Partners, billionaire Vinod Khosla, and others.
At the launch event, Bhavish Aggarwal showcased the capabilities of the Krutrim model, demonstrating its ability to generate text in different languages and contexts. He emphasized the importance of creating AI models that can capture the culture, language, and ethos of India, as most widely known AI models are trained primarily in English.
Additionally, Krutrim is actively involved within the Ola Group, utilizing its AI capabilities to enhance voice chat, sales calls, and customer support emails. The startup plans to release Krutrim Pro, an enterprise model, in the next quarter. Aggarwal revealed that he personally uses the software for writing performance reviews and composing job descriptions within his team.
Krutrim’s future plans include developing chips and employing a multiple chiplet strategy, which is expected to reduce costs and facilitate data center design.
This innovative endeavor by Krutrim aims to bridge the language gap and deliver AI solutions that cater specifically to the diverse Indian population. With the rapid advancements in language models and AI technology, India is determined to establish its position in the global AI landscape, simultaneously preserving its cultural and linguistic heritage.
