The inside story of building Padhify

Padhify started with a simple observation: preparing for an exam is not really one problem.

There is the problem of learning the syllabus. There is the problem of understanding what the exam actually asks. There is the problem of knowing what you personally do not understand. And there is the problem of deciding what to do with the limited time left.

Most preparation tools solve one of these problems at a time. We wanted to build something that could connect them. That became Padhify.

The first version of the product was much simpler than what exists today. But the idea behind it was already there: if AI could understand the actual exam, the material you are studying, and your own preparation, it could do more than answer questions. It could become part of the preparation itself.

Starting with the exam, not the chatbot

It would have been easy to start with an AI assistant. Instead, we started with the exam.

Competitive exams have years of history in their papers. The syllabus, question patterns, marking schemes, and recurring topics contain a huge amount of information about what preparation actually requires. We wanted that information to be usable.

So the early work went into building a structured collection of exam papers and making them searchable and usable by exam, subject, chapter, and year. That sounds less exciting than launching an AI feature. It turned out to be one of the most important decisions we made.

An AI system can be very good at explaining something. But if it does not understand the exam you are preparing for, it does not necessarily know whether that explanation is useful for you. We wanted Padhify to start with the source material.

The question we kept coming back to

As people started using the product, we noticed that the questions they asked us were often bigger than the product itself. They were not only asking for answers.

They wanted to know what to study next. They wanted to know whether a topic was actually important. They wanted to understand why they were losing marks. And they wanted to know whether they were making progress.

Those questions have something in common. None of them can be answered well from a single piece of information. You need the syllabus. The exam papers. The study material. The student’s attempts. The mistakes. The time remaining.

That changed how we thought about the product. We were not trying to build a better collection of study resources. We were trying to build the layer that connects them.

Building the foundation for an AI coach

This became particularly important as we started thinking about Pady. The obvious version of an AI tutor is reactive. You ask a question, it gives you an answer. We wanted something different.

If a student has an exam in three months, has already completed half the syllabus, keeps losing marks in the same three chapters, and has not revised another subject for two weeks, the useful question is not simply “what is the answer to this question?” It is “what should this student do next?” Answering that requires memory.

So before building the AI coach, we built the preparation record underneath it. Goals. Target exams. Dates. Attempts. Mistakes. Revision. Progress. The record changes as the student prepares.

That foundation is what eventually became Pady. Pady can use the history of a student’s preparation because that history already exists as structured information. It does not have to reconstruct the student’s situation from every new conversation.

Building Pady meant building more than an interface. Rohan Pundir, CEO, has been driving the product around the preparation problems we want Pady to solve, while Aahna S, our technical officer, has been building the technical foundation that brings those ideas together.

Making AI useful means giving it something to stand on

The next question was what Pady should actually know. General purpose AI has an enormous amount of knowledge, but exam preparation has a different requirement. The answer needs to be relevant to the exam, and ideally grounded in the material the student is using.

So we started building the other side of the system. Chapter material. Structured notes from lectures. Past year papers. Practice tests. Performance data. Each piece makes the others more useful.

A past paper tells us what an exam has asked. Study material helps explain it. An attempt tells us whether the student understood it. Performance tells us where the problem is.

Pady can use all of that context to decide what should happen next. That is the product we were trying to build from the beginning, even though it took time to make all the pieces exist.

Building with people who are actually preparing

A large part of Padhify has been shaped by talking to aspirants. Some feedback confirmed things we believed. Some completely changed our minds.

That distinction matters because building for exam preparation makes it easy to design around assumptions. You can look at how people are supposed to study and build a beautiful workflow around it. Actual students rarely follow that workflow.

They switch between lectures, notes, questions, tests, revision, and whatever they can fit into the time they have that day. So we have tried to build around the preparation that actually happens.

That is why Padhify is not just a content library or a test platform. It is becoming a system that connects learning, practice, performance, and planning.

Where Padhify is going

Padhify is still early. But the direction is clearer now.

We started by making exam material easier to access. We then built ways to practise it, understand it, study from it, and measure performance against it. Pady brings those pieces together.

The longer term idea is bigger than an AI tutor. We want Padhify to become a system that understands the exam, the material, and the person preparing for it well enough to make the next decision more useful.

There is still a lot to build. More exams. More material. Better ways to understand performance. Better ways for Pady to turn that understanding into useful action. That is what we are working on now.

The product is still taking shape, and the people using it are a big part of how we decide what comes next. If something you need is missing, we would rather hear about it now than build around its absence.