Onliversity: Personalized AI Learning

Inspiration

The inspiration behind Onliversity stemmed from a critical observation of traditional online learning platforms: they often fall short in delivering truly interactive, personalized, and engaging educational experiences. We noticed that many platforms rely on passive content, like pre-recorded videos or static text, which fails to replicate the dynamic nature of a one-on-one human tutoring session. This "one-size-fits-all" approach often leaves learners feeling disengaged, especially when tackling complex subjects or when they need immediate answers to specific questions. We recognized a significant opportunity to bridge this gap by leveraging the power of AI to create a new educational model that offers scalable, interactive, and personalized learning through a virtual teacher avatar.

What it does

Onliversity is an innovative, AI-powered e-learning platform that delivers personalized education through an interactive virtual teacher avatar. Our core offering is real-time, two-way voice conversations with an AI tutor that explains concepts, answers questions, and adapts its teaching style to the learner's comprehension level. Human educators curate and upload high-quality educational content (slides, notes, learning paths), ensuring academic rigor, while the AI dynamically delivers these lessons. This approach transforms passive learning into an engaging, conversational experience, making high-quality, personalized education accessible to a vast audience.

How we built it

We built Onliversity by integrating cutting-edge AI models with a robust web application stack. The frontend is developed using React.js, providing a dynamic and responsive user interface for both learners and teachers. For the backend and database, we leveraged Supabase, which offers a powerful and scalable solution for user authentication, data storage, and serverless functions.

The heart of our AI capabilities lies in the integration of two powerful models:

• DeepSeek AI: This model handles all our generative AI tasks. It's responsible for transforming raw content into professionally formatted lesson scripts, generating intelligent multiple-choice or true/false quizzes with explanations, creating interactive flashcards, and even designing personalized learning paths tailored to individual goals and knowledge levels. DeepSeek also aids in document parsing, extracting text from various file formats to facilitate content creation.
• Tavus CVI Agent: This is the engine behind our interactive AI tutors. Tavus creates realistic AI video avatars that deliver lesson content conversationally. It enables real-time, context-aware responses to student questions, simulating a genuine one-on-one tutoring session. Crucially, Tavus is integrated to understand a student's progress within a lesson, allowing the AI tutor to seamlessly pick up exactly where the student left off, ensuring a continuous and personalized learning journey.

Our platform flow is meticulously designed for both Learners and Teachers. Learners have a personalized dashboard with features like active courses, progress tracking, course recommendations, a personalized learning path, and a dedicated "Ask AI" page for general queries. Teachers, on the other hand, benefit from a comprehensive dashboard that enables course creation through a multi-step wizard, performance analytics, and profile management. The course creation process is significantly streamlined by AI-powered script, quiz, and flashcard generation, allowing teachers to focus on content quality and pedagogy.

Challenges we ran into

Building Onliversity presented several exciting challenges. One of the primary hurdles was integrating multiple advanced AI models (DeepSeek and Tavus) to work seamlessly together and deliver a cohesive, real-time interactive experience. Ensuring that the AI tutor could understand conversational context and respond accurately and dynamically to student queries, while also adapting to their progress, required meticulous API integration and prompt engineering.

Another significant challenge was designing a user interface that was both intuitive and powerful for two distinct user types: learners and teachers. We had to ensure that the learner dashboard provided a personalized and engaging experience, while the teacher dashboard offered robust tools for content creation and course management, all within a coherent design language. This involved careful consideration of information architecture and user flow for each role.

Finally, managing the scalability and performance of the platform, especially with real-time AI interactions and extensive content generation, was a continuous effort. Optimizing API calls, handling large datasets from content uploads, and ensuring a smooth user experience under varying loads required diligent attention to our Supabase backend and frontend optimizations. We also faced and resolved a bug related to accurate data retrieval for enrolled students in the teacher dashboard, ensuring reliable performance metrics.

Accomplishments that we're proud of

We are incredibly proud of several key accomplishments with Onliversity. First and foremost, we successfully created a truly interactive and personalized e-learning experience through our AI-powered virtual teacher avatar. The ability for students to engage in real-time, two-way voice conversations with an AI tutor, receiving immediate and tailored responses, is a significant leap forward from traditional passive learning.

We are also proud of the seamless Human-AI Collaboration model we've established. By allowing human teachers to focus on creating high-quality, structured educational content while the AI handles dynamic lesson delivery and scalability, we've created a powerful synergy that maximizes both academic rigor and accessibility.

The comprehensive and user-friendly dashboards for both learners and and teachers are another source of pride. We've designed intuitive interfaces that empower students with personalized learning paths and detailed progress tracking, while providing instructors with powerful AI-assisted tools for course creation, content generation, and performance analytics. The AI-powered quiz and flashcard generation features, in particular, significantly streamline the content creation process for teachers.

Finally, the robust integration of DeepSeek and Tavus CVI Agent demonstrates our technical prowess in leveraging cutting-edge AI to solve real-world educational challenges, making high-quality, personalized tutoring accessible at an unprecedented scale.

What we learned

Throughout the development of Onliversity, we learned invaluable lessons about the intersection of AI, education, and user experience. We discovered the critical importance of meticulous prompt engineering when working with generative AI models like DeepSeek to ensure the generated content (lesson scripts, quizzes, flashcards) is not only accurate but also engaging, well-formatted, and pedagogically sound.

We also gained a deeper understanding of the nuances involved in creating a truly conversational AI experience. It's not just about accurate responses but also about maintaining context, adapting to the user's comprehension level, and even incorporating non-verbal cues (through the Tavus avatar's body language and expressions) to enhance engagement and make the interaction feel more natural.

Furthermore, we learned the significance of designing for scalability from the outset. As an e-learning platform aiming for broad accessibility, ensuring our backend infrastructure (Supabase) could handle a large number of concurrent users and AI interactions was paramount. This involved continuous optimization and careful consideration of data models and API call efficiencies.

Lastly, the iterative process of gathering feedback and refining features based on user interactions proved essential. For instance, addressing the bug in the enrolled learners modal highlighted the importance of thorough testing and user validation to ensure all functionalities meet expectations.

What's next for Onliversity

For Onliversity, the future is bright and full of exciting possibilities. Our immediate next steps include:

• Expanding subject coverage and grade levels: While currently focused on secondary school, we plan to broaden our content to include middle school, college-level subjects, and professional upskilling courses.
• Enhancing AI adaptability and personalization: We aim to further refine the AI's ability to adapt to diverse learning styles, integrate more sophisticated assessment methods, and offer even more granular personalization in content delivery and learning path generation. This includes leveraging the "Student Profiler" for more tailored recommendations and quiz generation.
• Introducing collaborative learning features: We envision incorporating features that allow students to engage in group discussions with the AI tutor or collaborate on projects, fostering a more social learning environment.
• Developing a mobile application: To ensure maximum accessibility, we plan to develop native mobile applications for both iOS and Android, providing a seamless learning experience on the go.
• Exploring new monetization strategies: Beyond our current revenue streams, we will explore partnerships with educational institutions for white-label solutions, corporate training programs, and potentially a marketplace for third-party AI-generated content.
• Continuous improvement of AI models: We will stay abreast of the latest advancements in AI to continuously enhance the capabilities of our Tavus integrations, ensuring the most realistic and effective virtual tutoring experience possible.

---

Built with

• React.js: For building the interactive and responsive user interfaces for both learner and teacher dashboards.
• Supabase: Utilized as the backend and database solution, providing robust features for user authentication, data storage, and serverless functions, ensuring scalability and security.
• Tavus CVI Agent: Powers the interactive AI video avatars, enabling real-time, two-way voice conversations and dynamic content delivery based on student progress.

Built With

• openrouter
• react
• supabase
• tavus