Athena

Learn Better, Learn Faster
Architecture
Capabilities

💡 Inspiration

As students, we can relate to the difficulties of studying materials by ourselves alongside having to consume a mass amount of information. We also all have different learn styles, habits, and optimizations that we have developed over time.

In our research, we learned that students participating in peer discussions are TWICE as likely to answer conceptual questions correct.

Not only that, AI-personalized learning interventions led to learning gains 28% higher than control groups.

Thus, we considered how we might utilize AI-learning capabilities to reimagine learning through peer discussions while NOT impeding on our current learning processes that our tailored to each individual.

👀 What it does

We accomplish this by creating a Multi-Modal Experience of learning by providing different tool calls for AI to utilize in providing the best explanation to our user.

Athena is able to take in the context of all of your class' documents and from there generate summaries, practice quizzes, diagrams, animations, interactive demos, and more!

Furthermore, we also have access to a live desk view that you can interact and learn from! You can stream your hand written work at talk with Athena about your questions and ideas.

🛠️ How we built it

Utilizing Next.js, Vapi, and Google Gemini, we successfully created an Agent Workflow that allows you to converse with an AI Agent that will bring all the relevant information you would like while learning. This provides a hands-free, interactive learning experience that is personalized to you!

For Summaries and Quizzes, we utilized Google Gemini with our RAG system to generate relevant material to learn from.

For Diagrams, we utilized the Mermaid library to build beautiful diagrams from Flow Charts to Timelines to Mind Maps.

For Animations, we used Manim code that was generated by Google Gemini to effectively visualize complex topics.

For Interactive Demos, we had Google Gemini build out basic HTML, CSS, JavaScript Apps and embedded those to our platforms to have an interactive app to learn a concept from.

For PDFs and Videos that were uploaded, we utilized Google Gemini with RAG embeddings' metadata to pull from the relevant file source.

For Desk View, we live streamed our phone camera view to our web app and sent screenshots to Gemini to observe anything on the user's desk. This lets us ask questions and interact with our physical notes and assignments.

Challenges we ran into 🏃‍♂️

Some challenges we ran into was integrating the plethora of features and tools that we have separately implemented and also choosing the right models to create a seamless conversational AI experience.

We all pushed the bounds of what we knew; however, had a great time while doing it and learning together!

Accomplishments that we're proud of 🌟

We're proud of how seamless the user experience is with this multi-modal application. It works very well and as students, we can clearly see the use cases within our own studies.

What we learned

The entire team learned a lot from working together to technical up-skilling. This was our first time working with real-time conversational AI along with implementing many of the various custom tools that we built out from creating and simulating Manim code to live streaming video.

We all pushed the bounds of what we know and are very proud of our work.

What's next for Athena

We hope to begin tailoring the experience for every user and get the application production ready. We believe we are pushing the bounds of Human-Centered Interaction and reimagining learning workflows.