<p>Ola co-founder Bhavish Aggarwal has jumped onto the artificial intelligence (AI) models bandwagon with the launch of Krutrim, India’s first multilingual large language model (LLM). With this, he has joined the increasingly competitive AI race dominated by players such as Google, Microsoft, and OpenAI.</p>.<p>“The AI models known around the world are trained largely in English. They cannot capture our culture, language and ethos,” Aggarwal told the media in Bangalore, pursuant to the launch of his maiden AI product. “Our LLM is voice-enabled, and able to understand several languages and even as a mix of languages such as Hinglish — Hindi and English,” he added.</p>.<p>Trained by a team of leading computer scientists, based in Bangalore and San Francisco, the model powers a conversational AI assistant that can understand 22 Indian languages and speak (or generate output) in 10 of them. The chatbot will be available in a beta version in January.</p>.<p>Krutrim Si Designs, an AI venture, is an independent business, fully owned by Aggarwal with no plans of integration with his other businesses - including Ola Mobility and Ola Electric, he explained during a media roundtable. </p>.<p>Going forward, in Q4 FY24, the company will launch Krutrim Pro which will have multimodal, larger knowledge capabilities and many other technical advancements for inference. </p>.<p>The model has been built on the largest representation of Indian data used for its training, powering generative AI applications for all Indian languages. Data sources include a mix of proprietary data, data partnerships and open-source platforms, Aggarwal informed. </p>.<p>The company, which started work on this model a year ago, aims to launch Krutrim for both consumer applications and enterprise use cases. “We will also allow startups to build on our cloud technologies and our silicon (AI server board). So we will offer all - platform as a service, model as a service and software as a service,” Aggarwal added. </p>.<p>Aggarwal also revealed the company’s ambitious goal of launching an AI supercomputer in the next two years, for which it is developing data centers. Prototypes of the servers will be out in mid 2024 and production will start by the end of 2025.</p>.<p>The company is also focused on designing chips in what it calls the “multiple chiplet” strategy that it said will cut costs. It will then collaborate with semiconductor fabrication companies to manufacture these chips. </p>.<p>By March, Krutrim will be widely deployed within the Ola Group companies to aid voice chat, sales calls and customer support emails, Aggarwal said, adding that he uses the software to write performance reviews for his team and compose job descriptions for hiring.</p>
<p>Ola co-founder Bhavish Aggarwal has jumped onto the artificial intelligence (AI) models bandwagon with the launch of Krutrim, India’s first multilingual large language model (LLM). With this, he has joined the increasingly competitive AI race dominated by players such as Google, Microsoft, and OpenAI.</p>.<p>“The AI models known around the world are trained largely in English. They cannot capture our culture, language and ethos,” Aggarwal told the media in Bangalore, pursuant to the launch of his maiden AI product. “Our LLM is voice-enabled, and able to understand several languages and even as a mix of languages such as Hinglish — Hindi and English,” he added.</p>.<p>Trained by a team of leading computer scientists, based in Bangalore and San Francisco, the model powers a conversational AI assistant that can understand 22 Indian languages and speak (or generate output) in 10 of them. The chatbot will be available in a beta version in January.</p>.<p>Krutrim Si Designs, an AI venture, is an independent business, fully owned by Aggarwal with no plans of integration with his other businesses - including Ola Mobility and Ola Electric, he explained during a media roundtable. </p>.<p>Going forward, in Q4 FY24, the company will launch Krutrim Pro which will have multimodal, larger knowledge capabilities and many other technical advancements for inference. </p>.<p>The model has been built on the largest representation of Indian data used for its training, powering generative AI applications for all Indian languages. Data sources include a mix of proprietary data, data partnerships and open-source platforms, Aggarwal informed. </p>.<p>The company, which started work on this model a year ago, aims to launch Krutrim for both consumer applications and enterprise use cases. “We will also allow startups to build on our cloud technologies and our silicon (AI server board). So we will offer all - platform as a service, model as a service and software as a service,” Aggarwal added. </p>.<p>Aggarwal also revealed the company’s ambitious goal of launching an AI supercomputer in the next two years, for which it is developing data centers. Prototypes of the servers will be out in mid 2024 and production will start by the end of 2025.</p>.<p>The company is also focused on designing chips in what it calls the “multiple chiplet” strategy that it said will cut costs. It will then collaborate with semiconductor fabrication companies to manufacture these chips. </p>.<p>By March, Krutrim will be widely deployed within the Ola Group companies to aid voice chat, sales calls and customer support emails, Aggarwal said, adding that he uses the software to write performance reviews for his team and compose job descriptions for hiring.</p>