TutorSaathi AI: An Open-Source AI Teaching Assistant for English Tutors 🇮🇳

TutorSaathi AI is an open-source AI teaching assistant created to assist English tutors in generating personalized learning materials for Hindi-speaking students. Despite students' understanding of English grammar, they often find it challenging to apply those rules in real-life conversations. Meanwhile, tutors invest considerable time in crafting worksheets, quizzes, translations, and practice exercises for their pupils. TutorSaathi AI seeks to streamline this process, making it quicker and more accessible for educators.

Key features of TutorSaathi AI include translating Hindi sentences into natural English with concise explanations and examples, generating grammar exercises, vocabulary activities, quizzes, and answer keys, creating personalized practice based on a student's learning level and common errors, providing daily English conversation prompts to build learners' confidence, and tracking progress by reviewing quiz results and identifying areas requiring additional practice.

The target audience for TutorSaathi AI comprises independent English tutors, small coaching centers, and online educators. This tool aims to save preparation time and offer more personalized learning experiences. The project's long-term goal is to deliver a valuable free learning experience while providing affordable premium features for tutors who require additional worksheets, student management, and progress reports.

The demo will feature a tutor generating an English worksheet, explaining Hindi sentences in English, and designing practice activities for students. The project's source code, setup instructions, and documentation will be accessible in the repository.

TutorSaathi AI employs a full-stack web application approach with a user-friendly interface for students and teachers. The technology stack comprises Next.js and TypeScript for the frontend, Tailwind CSS and shadcn/ui for UI, Gemma, an open-weight model family from Google, as the AI model, A compatible inference runtime for generating educational content as the inference layer, MongoDB Atlas for database management if persistent user accounts and progress tracking are implemented, and Vercel for web application deployment.

The core AI workflow involves using Gemma to generate translations, grammar explanations, worksheets, and practice questions based on user-provided topics and sentences. The application will guide the model using structured prompts and validate the generated output before presenting it to learners. Educators should carefully review AI-generated explanations, as they can sometimes be incorrect. The implementation and final technology choices will be documented in the GitHub repository.

Open innovation plays a crucial role in making educational tools accessible to learners and teachers, regardless of their budget. Leveraging an open-weight model like Gemma offers several benefits, such as greater flexibility in experimenting with different model sizes and inference setups, deployment choices depending on hardware and model requirements, transparency and experimentation by allowing developers to inspect model documentation, evaluate outputs, and enhance the application around the model's strengths and limitations, cost control through self-hosting or selecting an appropriate inference provider, and privacy-conscious design by reducing the need to send educational content to an external model provider.

However, using a closed API can also be valuable, and an open-weight approach provides developers with more control over experimenting, deploying, and adapting their AI-powered products.

TutorSaathi AI aims to explore how open-source AI can facilitate personalized English learning, making it more affordable and practical. The developer plans to test the application with an English tutor or a small coaching center and gather feedback on worksheet quality, usefulness, and time saved during lesson preparation. The feedback will help identify useful features and areas needing improvement.

The next priorities for TutorSaathi AI include testing the generated educational content, enhancing personalization, collecting feedback from real tutors, and exploring an affordable pricing model. The ultimate goal is to create a practical teaching tool that addresses a real problem for educators and their students. Open-source AI combined with practical education and real user feedback can result in a product worth building.