Ex-Google VP's AI Tutor Makes Learning Harder to Make Students Smarter

SAN FRANCISCO, CA – January 23, 2026 – In an era where artificial intelligence is often synonymous with instant answers, a new venture led by a former Google and Airbnb executive is taking a deliberately counterintuitive approach. Fermi.ai, an AI-first edtech startup, launched its platform today in the United States and India with a bold mission: to make learning harder. The company, co-founded by ex-Google GM & VP Peeyush Ranjan, aims to restore what educators call 'productive struggle' in high-school STEM education, believing the path to mastery isn't paved with shortcuts.

Headquartered in Singapore and emerging from the Meraki Labs ecosystem, Fermi.ai is the brainchild of Ranjan and his partner Mukesh Bansal, the entrepreneur behind Indian success stories Myntra and Cure.Fit. The platform debuts with a focus on Math, Physics, and Chemistry, targeting the lucrative after-school tutoring markets in India ($8 billion) and the US ($5 billion). Instead of serving as an answer machine, Fermi.ai is designed to act as a personal tutor that guides, probes, and challenges students, ensuring their brains stay engaged in the reasoning process.

"The industry has spent years building AI that gives answers," said Peeyush Ranjan, CEO of Fermi.ai. "We built Fermi.ai to do the opposite: to protect the 'productive struggle' that leads to actual mastery. We want to use AI to keep the brain working, not give it a reason to switch off."

The AI Paradox: A Coach Against Shortcuts

Fermi.ai enters an edtech market grappling with an "AI tutor race" that has simultaneously created a credibility problem. While many platforms promise personalized learning, their utility is often undermined when students use them to simply copy solutions, bypassing the cognitive effort required for deep understanding. Fermi.ai positions itself as an "anti-shortcut" tool, designed to combat this trend.

The platform's philosophy is rooted in the pedagogical principle that grappling with difficult concepts is essential for building robust neural pathways and true comprehension. It's a direct challenge to the culture of immediate gratification fostered by general-purpose AI models. Where a student might ask a standard chatbot for the solution to a physics problem and receive it instantly, Fermi.ai's adaptive tutor would instead ask a clarifying question, suggest a different approach, or prompt the student to re-examine a specific concept, mimicking the Socratic method of an expert teacher.

This approach is built on a sophisticated technical foundation that leverages multiple frontier AI models, including GPT 5.2 and Gemini 2.5. The company uses a proprietary benchmarking tool to select the best-performing model for any given problem type, ensuring it remains agile and effective. The goal is not to replace student thinking but to support and scaffold it.

Reconnecting with Reason Through Digital Ink

The platform's commitment to deeper learning is most evident in its four core pillars, particularly its unique user experience. The first is its Adaptive Real-time Tutor, which provides stepwise coaching instead of final answers.

Second, and perhaps most distinctively, is the Handwriting First Experience. Recognizing that STEM problem-solving is a tactile process involving equations, free-body diagrams, and molecular structures, Fermi.ai prioritizes digital ink and stylus input on a smart canvas. This design choice aims to eliminate the clunky process of "prompt engineering" and allows students to think and work naturally, as they would with pen and paper. This feature is designed to be accessible on a range of devices, from iPads to more affordable Android tablets.

The third pillar is a Concept Graph and Question Bank, which maps out curricula for exams like the AP, IB, and JEE. This system ensures the platform can always serve the "next right problem" to keep a student perfectly balanced in the zone of productive struggle. Finally, a Diagnostics & Analysis layer gives both students and educators unprecedented insight. For teachers, a feature called 'Classroom Command' surfaces exactly where a student's logic faltered, turning what was once 'silent struggle' into an opportunity for targeted intervention.

From E-Commerce Giants to EdTech Disruption

The ambition behind Fermi.ai is backed by some of the most experienced minds in scaling global technology platforms. Peeyush Ranjan's career spans leadership roles at Google and Airbnb, as well as serving as CTO for the Walmart-owned e-commerce giant Flipkart. His partner, Mukesh Bansal, is a celebrated entrepreneur who co-founded fashion retailer Myntra and later served as Head of Commerce at Flipkart before launching the health-tech platform Cure.Fit and the venture studio Meraki Labs.

This background in building and scaling massive, consumer-facing products informs Fermi.ai's strategy. The founders are applying lessons from e-commerce and large-scale systems to solve a fundamental problem in education that has been exacerbated by the very technology they helped pioneer. Their focus is on creating a product that not only works but can also scale globally to redefine AI's role in the classroom.

"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," added Mukesh Bansal, Partner at 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."

Measuring Mastery and the Path Ahead

Before its public launch, Fermi.ai conducted a three-month pilot with 79 high-school students, yielding promising initial data. The results revealed what the company calls a 'Mastery Curve,' showing clear progress. Students who initially scored poorly on a concept (2 out of 10 or less) saw their scores improve by an average of 4.68 points by their final attempt. Across all subjects, overall mastery scores rose by 2.6 points from the first 20% of practice attempts to the last 20%.

Significantly, students who engaged most heavily with the platform (completing 100 or more practice attempts) not only saw their scores improve by 2.82 points but also reduced their reliance on hints by 21%, suggesting a tangible increase in independence and understanding. Qualitative feedback from the pilot indicated students appreciated the 'judgment-free' coaching environment.

However, the company is transparent about the nature of these findings. A company whitepaper notes that these mastery grades are internal platform measures, not scores from externally administered standardized tests. While the signals are encouraging, they represent a starting point that will require more robust, independent validation. To that end, Fermi.ai is now launching 12-16 week controlled pilots using cluster randomization to measure gains against external assessments and curriculum-aligned growth.

As of today, the cloud-based platform is available for students, while educators and schools are invited to join the 2026 pilot program to further test its impact in a classroom setting.