Mukesh Bansal, Peeyush Ranjan team up to launch AI-first edtech start-up

At a time when artificial intelligence (AI) is making answers instantaneous but learning increasingly superficial, Mukesh Bansal, former Founder of Walmart-owned Myntra, along with Peeyush Ranjan, former GM & VP at Google and Airbnb, and former CTO of Walmart-owned Flipkart, are betting on a different promise: AI that prompts students to think.

The duo, also Partners at Meraki Labs, have founded Fermi.ai, an AI-first edtech start-up focused on high-school STEM education, with initial rollouts in India and the US. Headquartered in Singapore, the company operates through subsidiaries in Bengaluru and the US and is part of the Meraki Labs ecosystem.

Unlike popular AI chatbots that generate solutions on demand, Fermi.ai is designed to guide students through the process of problem-solving rather than delivering final answers. The platform mimics the behaviour of a teacher—prompting students with questions, identifying faulty reasoning, and offering hints only when required.

“If a student says, ‘I don’t know what to do,’ a good teacher won’t jump to the solution. They’ll ask them to read the problem again, identify variables, draw a free-body diagram,” Ranjan said in an interaction with Business Today. “That is the behaviour we’ve modelled into the AI.”

The platform runs entirely on AI, with no human tutors in the loop. Built on multiple large language models and delivered via a stylus-first digital canvas, Fermi.ai tracks how students approach each step, flags errors in real time, and adapts prompts based on individual struggle patterns.

“We have entered an era where AI can solve any equation, but it can’t yet explain why a student’s logic faltered at step three,” said Mukesh Bansal, Partner Meraki Labs. “Fermi.ai isn't here to give answers; it's here to provide the map and the mirror - showing students how they think and giving teachers the visibility to lead them back to the path of mastery.”

In pilots conducted in Bengaluru, North India and Silicon Valley, Fermi.ai recorded an abandonment rate of just 13–15%. More significantly, students who initially scored poorly—around 2 out of 10—improved to an average of 6.7 over repeated attempts.

“What we’re taking away is the fear of getting stuck,” Ranjan said. Fermi.ai is initially targeting students from grades 8 to 12, starting with mathematics, physics and chemistry, but the underlying platform can be extended to new subjects within weeks. The team has already tested use cases in areas such as quantum physics and microeconomics.

The company is currently offering the platform as a free public beta while it finalises its revenue model, which is expected to be subscription-based with tiered usage and top-up options. Institutional partnerships with schools and educators are also part of the roadmap.

With a 10-member, entirely tech-focused team and no external funding so far, Fermi.ai plans to explore capital raising once pricing and go-to-market strategies are finalised. For Ranjan, the timing is deliberate. “In a post-AI world, every industry has to rethink from first principles,” he said. “Education cannot be an exception.”