AI unicorn Sarvam has appointed Devendra Singh Chaplot, a founding team member at France-based Mistral AI and former OpenAI executive Mira Murati’s Thinking Machines Lab, as an advisor as the startup steps up its push to build frontier AI models in India.
Chaplot’s appointment was announced today at Sarvam’s first developer conference, Epoch 2026, alongside a slew of launches across agentic AI, speech, vision and enterprise productivity.
Prior to this, Chaplot served as the pre-training lead at Elon Musk-led xAI. He has also worked on training several large language models, including Mistral 7B, Mixtral 8x7B and Mistral Large.
His appointment comes as the AI startup plans to build a trillion-plus parameter frontier AImodel from scratch in India over the next six months. Sarvam is also building an end-to-end AI stack spanning models, infrastructure and enterprise software.
The Bengaluru-based AI unicorn also announced the opening of a San Francisco office as part of its expansion.
Speaking at the event, Chaplot said building frontier-scale models in India is achievable with the right compute infrastructure. According to him, a model with around 3 Tn total parameters and roughly 100 Bn active parameters could be pre-trained in about two months using 10,000 Blackwell GPUs, while the entire training cycle, including post-training and reinforcement learning, could take around four to six months.
Chaplot also pushed back against the view that frontier AI labs require large teams of highly experienced researchers. He said a handful of experienced people, supported by motivated and talented teams, could build the required expertise relatively quickly.
“I don’t think experience is the bottleneck. If you have talented, innovative people, this can be built,” he said.
He also argued that India needs to develop the capability to train its own models rather than depend entirely on a handful of global AI systems, particularly given the country’s linguistic, cultural, educational, legal and public policy requirements.
“The model is not the goal.The ability to build models is the goal,” Chaplot said, adding that developing expertise across research, infrastructure and data would allow companies, governments, developers and researchers in India to build and customise models for their own needs.
The appointment comes as Sarvam looks to strengthen its frontier AI capabilities across coding, cybersecurity, simulation and scientific workloads.
Founded in 2023 by Pratyush Kumar and Vivek Raghavan, Sarvam develops AI models, infrastructure and enterprise products across language, speech, vision and document intelligence.
The startup began with OpenHathi-v1 in December 2023 and later launched smaller Indic language models such as Sarvam-2B and Sarvam-1.
It subsequently moved to larger models with Sarvam-30B and Sarvam-105B, which are designed for enterprise reasoning, multi-turn conversations and long-document understanding across more than 22 Indian languages.
Its AI products are already being used across banking, insurance, government and defence. Sarvam says its conversational AI platform handles more than 2 Mn interactions daily, while its inference platform processes over 10 Mn API calls a day.
The latest announcements come about a month after Sarvam became India’s 130th unicorn after raising $234 Mn as part of its $300 Mn Series B round at a post-money valuation of $1.5 Bn.
The round was led by HCLTech, which committed $150 Mn, and also saw participation from Bessemer Venture Partners, Khosla Ventures, and Peak XV Partners.
Sarvam plans to use the capital to develop next-generation models focused on agentic AI, coding and cybersecurity, while expanding its compute infrastructure and enterprise deployments.
As per unaudited financials shared by HCLTech, Sarvam clocked revenue of ₹45.1 Cr in FY26, up from ₹1.5 Cr in FY25.
The post Sarvam Ropes In Mistral AI Founding Team Member Devendra Chaplot As Advisor appeared first on Inc42 Media.
