StudySaathi AI 2.0: I Built a Local AI Study Companion for a Friend Who Wanted Learning to Feel Easier

Meghana Bobbili — Sun, 04 Oct 2026 09:43:17 +0000

This is my submission for the Hacktoberfest Weekend Challenge: Build for a Friend.

Most of the projects I have built before were for an imaginary "user". This time, the challenge asked me to build for a real person, so I started with a different question:

"How can I make studying easier and more manageable for my friend?"

The answer became StudySaathi AI 2.0 (స్టడీ साथी), a local AI study companion for school and college students. Saathi means "companion", and that is the feeling I wanted: something that sits next to you while you study, explains things simply, and helps you revise.

The AI runs locally through Ollama with llama3.2:latest.

What I Built

StudySaathi AI 2.0 is a local AI study companion designed around a real problem faced by my friend.

My friend has difficulty with a few things that, put together, make studying slower and more tiring than it should be:

English spellings are hard for her.
Difficult concepts can be hard to understand, especially when the explanation is complicated.
Some subjects and topics take more effort for her to handle than others.
Learning difficult topics takes her longer.
Revising topics quickly is also a challenge.

I didn't want to exaggerate her problem. She is not lacking effort. She needs simpler explanations, more time to go step by step, and an easier way to revise.

So instead of building a generic chatbot, I built features around those needs.

AI Study Buddy

This is the heart of the project. A student can ask about a concept by typing or using voice and choose how they want it explained:

Explain (Beginner) — simple explanations
Teach Step-by-Step
Real-World Example
Exam Mode (10-Marker) — 10-mark answer style
5-Point Quick Revision

The student can switch between English, Telugu and Hindi, use voice input, and listen to the answer through audio readout.

The Study Buddy can also show flowcharts and diagrams for concepts using Mermaid.js.

Adaptive Quizzes

The quiz section helps students practise topics, track mastery, and identify topics that need more practice.

Flashcards

Flashcards provide quick revision. Students can mark cards as Needs Practice or Mastered, and can also use the Read Question option for audio.

Smart Study Planner

The student can enter syllabus topics, the number of days remaining, and available study hours.

The planner organizes study into:

Learn
Practice
Quiz
Revision

Notes to Questions

Students can paste study notes or provide PDF/text input and generate questions for practice.

Code Playground

For college students, the project includes educational examples such as:

Linear Regression
Decision Trees
Binary Search
Quadratic Solver

Dashboard and Focus Tools

The dashboard brings study progress and planning together and includes a Pomodoro/focus timer to support revision.

Demo

I recorded a screen demonstration of the working project.

▶️ Watch the StudySaathi AI 2.0 Demo

The video shows StudySaathi AI 2.0 in action.

Code

The complete source code is available on GitHub:

GitHub Repository — StudySaathi AI 2.0

How I Built It

The architecture is intentionally simple:

Student → Frontend → Flask Backend → SQLite + Local Ollama AI

Frontend

HTML5
CSS3
Vanilla JavaScript / ES6
Web Speech Recognition
Audio readout
Mermaid.js for diagrams

Backend

Python
Flask
Flask-CORS
PDF and text processing

Database

SQLite

AI

Ollama
llama3.2:latest

The browser communicates with the Flask backend. Flask handles the application logic and database operations and communicates with the local Ollama model when AI is required.

In simple terms:

Student → Browser → Flask → Local Ollama → Response → Browser

Why Does Open Innovation Matter?

StudySaathi AI 2.0 uses Ollama + llama3.2:latest, with AI inference running locally instead of relying on a paid external AI API for the core AI functionality.

This mattered to me for several reasons:

The interaction can stay on the student's own machine when the required model and hardware are available. For educational interactions, I think that is a privacy-conscious choice.
There is less dependence on third-party cloud AI services for the core AI experience.
No paid external AI API is needed for the core AI features.
I have more control over the AI setup. As a developer, I can decide how the model is connected and used.
It is a learning opportunity. Local AI helped me understand how an AI model can be integrated into a complete application rather than treating AI as just an external API.

An honest limitation

Local AI depends on the user's computer resources, and smaller local models may not provide the same capabilities as very large cloud models.

I chose local AI because control, privacy and accessibility mattered for this project, not because I think it replaces large cloud models in every way.

Building It for a Real Friend

I did not start by asking:

"What generic AI app can I build?"

I started with:

"What problem does my friend actually have?"

Every major part of StudySaathi comes from one of her needs:

My friend's need	What I built
Difficult concepts	Simple, beginner-friendly and step-by-step explanations
Topics that take longer to understand	Guided learning and a study planner
Needing to revise	Quick revision mode, flashcards and quizzes
English difficulties	English, Telugu and Hindi support
Needing another way to interact	Voice input and audio readout

Because of this, the project was shaped by a real person rather than only by a technology idea.

When I was unsure whether a feature belonged, I could ask myself:

"Would this make studying easier for her?"

What My Friend Said

My friend tried StudySaathi AI 2.0.

Her feedback was that she feels her concepts and subjects are easier to handle with it, and that she found it helpful for quick revision.

That is the feedback I have, and I'm sharing it as it is. I'm not claiming anything beyond it.

What I Learned

Building StudySaathi taught me several things:

Building around a real person's problem is different from building a generic project. The requirements came from her needs, so deciding what to build became much clearer.
Local AI integration means understanding how the application talks to the model. It is not just "add AI". The frontend, backend and model all have to work together.
AI features are only useful if they actually help the user. A feature that looks impressive but doesn't help my friend study is not worth much.
A complete application connects frontend, backend, database and AI. Making all four work as one product taught me more than building any single part.
Feedback from a real friend shows whether a feature is useful. Her response told me more than my own testing could.

What's Next

These are plans, not finished work:

Better spelling-learning support
More personalized revision
More local/open models
Easier installation and setup
Better multilingual learning
More real-user feedback
Additional study analytics

Final Thoughts

StudySaathi AI 2.0 is not intended to replace teachers or solve every learning difficulty.

It is a study companion designed to help students understand, practise and revise topics more easily.

"Build for a Friend" changed the way I think about building software. Instead of beginning with a model or a framework, I began with a person and one question:

"How can I make studying easier for her?"

The answer was a local AI companion that explains simply, helps with revision, supports multiple languages and provides another way to interact.

Hearing that she finds her concepts and subjects easier to handle, and that it helps her revise quickly, means more to me than any technical detail.

If you want to build something, I'd suggest starting with someone you know and a problem they really have.

Thanks for reading!

devchallenge #weekendchallenge #hf26challenge

DEV Community: Meghana Bobbili