Wingify founder Paras Chopra is recruiting to build India’s foundational AI lab

Representative image

Paras Chopra, who recently exited Wingify in a $200- million deal with Everstone Capital, is now assembling a team to build a foundational artificial intelligence (AI) lab in India.

"I'm assembling a team to put a proposal for India's Al mission if you're interested in joining the team full-time or part-time," Chopra announced on the X platform, formerly Twitter.

On January 30, the government announced plans to develop a generative AI model, looking to join the ranks of global players such like OpenAI's ChatGPT and China's DeepSeek, which recently took the tech world by storm with its low-cost AI assistant.

The initiative would be powered by the IndiaAI Compute Facility, which has secured 18,693 GPUs to drive the development of a Large Language Model (LLM) tailored specifically for India, IT minister Ashwini Vaishnaw said.

It was decided to launch a foundational AI model after India Inc and other AI experts debated whether the country should build its own LLM.

Vaishnaw said proposals for AI model development would be invited starting January 30, with an expected development timeline of four to eight months.

Also read: DeepSeek's LLM success triggers big debate: Is India's hesitation a strategic mistake?

Paras said hard and complex problem solving is the bedrock of all economic growth. "If we crack super-intelligent reasoning models that think creatively, we will turbocharge the growth and have a real chance of solving many of the world's current unsolved problems like poverty, cancer, and the mysteries of universe and consciousness," Chopra wrote on the social media platform.

He listed three reasons why focus on efficiency is important. Firstly, unlike western Al labs, India does not have a billion dollars to spend on GPUs at the start. Secondly, this will force the country to look at problems from a fresh perspective, as constraints will lead to reasoning creativity. Lastly, Chopra said he is inspired by the efficiency of the human brain which has made itself energy and data-efficient.

"Within reasoning, we have a lot of attack vectors for efficiency: e.g. explore tools, RL, distillation, pruning, representations, state space models for linear time inference, neurosymbolic reasoning, test time learning, etc," he wrote.