Bhavish Aggarwal, the co-founder and CEO of Ola, recently launched a new venture called Krutrim, focusing on providing artificial intelligence (AI) solutions for the Indian market. One of its notable offerings is the Krutrim AI chatbot, capable of understanding prompts in 22 Indian languages and responding in English, Hindi, and eight other Indian languages.
Aggarwal shared that Krutrim aims to develop a comprehensive AI stack, starting from semiconductor and GPU levels to creating Large Language Models (LLMs). The startup plans to invest in chip designs, sustainable cloud architectures, and innovative models for various sectors like healthcare, drug discovery, education, agriculture, and more. Additionally, Krutrim aims to create consumer-facing applications.
Importantly, Krutrim operates independently of Ola and Ola Electric, though there may be some undisclosed investments from these entities. While Aggarwal did not disclose the exact investment size, reports suggest that Krutrim secured $24 million in debt from Matrix Partners.
The Krutrim AI chatbot stands out by supporting a wide range of Indian languages, distinguishing itself from well-known AI chatbots like ChatGPT and Google’s Bard. It has generative capabilities across 10 Indian languages and can process inputs in a total of 22 languages. The system underwent training on a massive dataset containing over two trillion tokens specific to Indian languages.
Aggarwal emphasized the importance of training AI models in languages that reflect cultural values, context, and ethos. He expressed the need for India to lead in AI by training models on unique datasets specific to the country and ensuring accessibility with cost-effective structures.
Krutrim plans a phased release, initially providing access to those who signed up, with a public release scheduled for January 2024. Open APIs will be accessible to developers by February 2024. Ola intends to integrate Krutrim across its group of companies for various functions such as sales, service, support, and operational processes.
Krutrim is actively involved in developing AI infrastructure, including indigenous data centers, server computing, edge computing, and supercomputers. Prototypes are expected by mid-2024, with a production roadmap set for release by the end of 2025. Aggarwal highlighted the importance of designing their own silicon chips for efficient, cost-effective, and sustainable data centers, with the expectation of having their own chip by 2025.
