TutorKitAI: Building a Local AI Teaching Assistant with Ollama

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

I built TutorKitAI, a small AI-powered teaching assistant that helps create study materials without having to write everything from scratch.

I got the idea from a simple question: what if a tutor could enter a topic, choose a grade and language, and get a useful starting point for a lesson or worksheet in a few seconds?

That is what I wanted to try with this project.

TutorKitAI can generate:

• Lesson plans for different subjects and grades
• Worksheets with practice questions and answer keys
• Quizzes for checking understanding
• Revision notes for exam preparation
• Simple explanations of difficult topics

I also added language options for English, Telugu, and Simple English. I wanted to include Telugu because learning materials should be useful to students who are more comfortable learning in their own language.

The idea is to make preparation a little easier for tutors, teachers, and people who teach part-time. Instead of starting with a blank page, they can generate a draft and adapt it for their students.

For the side-hustle angle, I can see this being useful for independent tutors and small tuition centres that need to prepare learning materials regularly. It could eventually become a paid tool, but right now I'm focusing on getting the core features working properly.

Demo

Project: TutorKitAI

GitHub repository:

The project runs locally on my computer. I haven't deployed a public live demo yet, so the repository is the best place to explore the code.

Code

25A35A0530 / TutorKitAI

TutorKit AI ✳

A local AI-powered teaching assistant that helps tutors create educational materials using Ollama.

TutorKit AI helps tutors prepare lesson plans, worksheets, quizzes, revision notes, and simple explanations through an easy-to-use web interface. It runs locally on your computer using an Ollama model, allowing you to generate educational content without depending on a hosted AI API.

✨ Features

• Lesson Plan Generator — Create structured lesson plans with objectives, teaching steps, activities, assessment, and homework.
• Worksheet Generator — Generate practice questions with a separate answer key.
• Quiz Generator — Create multiple-choice questions with answer keys and explanations.
• Revision Notes — Generate topic summaries, definitions, examples, and practice questions.
• Simple Explanations — Break down topics into short, student-friendly explanations.
• Multilingual Output — Generate content in English, Telugu, and Simple English.
• Difficulty Selection — Choose Easy, Medium, or Hard.
• Grade Selection — Tailor content to the selected grade or learner level.
• …

View on GitHub

The repository contains the frontend, Python server, and README with setup instructions.

How I Built It

I wanted to keep the project simple enough to understand and run on my own computer, so I went with a lightweight stack.

Here's what I used:

• Python for the backend server
• HTML, CSS, and JavaScript for the interface
• Ollama to run the AI model locally
• Qwen2.5-Coder 3B as the model for generating educational content
• Local storage to keep saved materials in the browser

The frontend sends a request to the Python server, which communicates with Ollama's local API. The generated content is then displayed in the interface, where users can copy it, download it, or save it to their library.

One thing I liked about this setup is that I didn't need to build the whole project around a paid AI API. The model runs locally, which made experimenting with prompts much easier.

Testing was an important part of the process. I tried different topics and content types, including a lesson plan on the water cycle, fraction worksheets, and simple explanations in Telugu.

The Telugu output also reminded me that getting a response from an AI model isn't the same as getting a good response. I had to pay attention to repetition, clarity, and whether the generated content actually followed the prompt. There is still room to improve the quality, especially for educational content in different languages.

I also tested basic features such as copying, downloading, and saving generated materials. I want the app to be useful beyond the moment when the AI finishes generating text.

Why Does Open Innovation Matter?

For me, open innovation made it possible to experiment without making API costs the first obstacle.

Using an open-weight model through Ollama meant I could run the model on my own machine, test different prompts, and build the surrounding application myself. I could work on the frontend and backend while keeping control over how the AI was used.

It also made the project easier to experiment with at my own pace. I didn't need to depend on a hosted AI service for every test.

Of course, running a model locally has its own limitations. Output quality depends on the model, and smaller models may struggle with complex instructions or some languages. Local inference also depends on the computer's available resources.

But that trade-off was worth exploring for this project. I got to learn how an AI-powered application works from the interface to the model response, instead of only connecting to an external API and treating the model as a black box.

My Agent Session

I haven't published a DevRelay agent session for this project yet, so I'm leaving this section out rather than linking to something that isn't available.

Prize Categories

Side Hustle — TutorKitAI explores how a local AI tool could help independent tutors and small tuition centres spend less time preparing learning materials.