/ciol/media/agency_attachments/c0E28gS06GM3VmrXNw5G.png)
Follow Us/ciol/media/media_files/2026/01/23/fermi-2026-01-23-21-05-59.png)
Former Google executive Peeyush Ranjan, along with entrepreneur Mukesh Bansal, has launched an AI-driven edtech startup that wants to teach students how to think, not just find answers faster.
Fermi.ai, based in Singapore with operations in India and the US, takes a different approach to AI tutoring. Instead of spitting out solutions, the platform walks students through problems step-by-step in math, physics and chemistry, basically STEM subjects.
"Students today are getting answers faster than ever, but their understanding is getting weaker," Ranjan said. "We built Fermi.ai to support thinking, not replace it."
The platform is currently free at fermi.ai. The team plans wider rollouts and school partnerships later in 2026, with pricing models still under discussion. Ranjan said they're considering per-student or per-seat fees once the product matures.
Ranjan previously served as general manager and vice president at Google, and later as CTO of Flipkart. He co-founded the startup through Meraki Labs, an incubator he runs with Mukesh Bansal, who co-founded Myntra.
The platform is built around what Ranjan calls "productive struggle"—letting students work through challenges with guidance rather than instant answers. It includes a real-time adaptive tutor, a digital canvas where students can write equations by hand, exam-aligned practice questions for AP, IB and JEE, and analytics that show teachers exactly where students get stuck.
Bansal said the focus is understanding thought patterns, not just speed. "It's about showing students how they think, and helping teachers guide them back to mastery," he said.
Before launch, Fermi.ai ran a three-month pilot with 79 students across more than 15,000 practice problems. Students who initially struggled showed improved mastery scores and used hints less over time, according to the company.
As AI tutoring tools flood the edtech market, from ChatGPT-based homework helpers to personalized study apps, Fermi.ai is betting on depth over shortcuts. The pitch is real learning takes longer than a quick answer, and AI should make that struggle more effective, not eliminate it entirely.