Tutor.AI

pipeline diagram
user intereraction

Inspiration

Ever felt the struggle of tackling those never-ending educational videos? That's the inspiration behind Tutor AI. We get it, and we're here to shake up the learning game by making the whole process way easier and way more fun.

What it does

Tutor.ai leverages advanced natural language processing and video analysis to transform the way you engage with educational content. Our cutting-edge technology parses and processes all aspect from video lectures, from text to vision. Simply ask your question, and Tutor.ai not only provides you with accurate answers but also guides you to the precise timestamp in the video where the relevant information is discussed. With Tutor.ai, learning becomes efficient, interactive, and tailored to your needs.

How we built it

Please refer to the pipeline diagram and technology we used below. In essence, we dissect a video input into various information sources: transcriptions from audio, slide details, and facial expressions from keyframes. We then ingest this wealth of data into the vector database to glean deeper insights about the video, thereby enhancing our ability to provide more informed responses to users' inquiries.

Challenges we ran into

Precision in Timestamps: Ongoing efforts to refine our parsing algorithm for accurate timestamps in diverse video and lecture formats.
Integration Complexity: Continuously addressing challenges in integrating Tutor.ai with varied sources, refining our architecture to overcome format obstacles.
Data Standardization: Iterative approach to standardizing diverse data formats in videos and lectures through meticulous preprocessing.
Optimizing for Scalability: Constant considerations and improvements in scaling Tutor.ai for diverse content formats.

Accomplishments that we're proud of

-Develop a user-friendly interface for easy video URL input and seamless navigation between the chatbox and video. -Establish a robust backend system using AstraDB for efficient video importation and storage. -Implement the RAG cycle to accumulate knowledge from imported videos, enhancing response accuracy. -Leverage AstraDB and Llama-index for scalable and efficient knowledge accumulation and retrieval.

What we learned

RAG Llama-index AstraDB How to build robust end-to-end pipeline

What's next for Tutor AI

Extend support beyond videos and lectures to include additional input sources such as ppt, pdf, audio recordings, and more.
Invest in refining parsing algorithms for even greater accuracy in pinpointing relevant moments within diverse content.
make it a platform. Explore features that foster collaborative learning experiences, allowing users to engage and share insights.
Work towards making Tutor AI accessible on mobile devices, ensuring learning is convenient and on-the-go