The Story of Grasphoffer: Leaping Over Learning Hurdles
Our Inspiration
As students, we've all felt the frustration of hitting a wall in our studies because we were missing a key piece of foundational knowledge. This "prerequisite knowledge gap" is a universal struggle. Our inspiration was solidified by Benjamin Bloom's 1984 "2 Sigma Study," which proved that students with one-on-one tutoring and mastery learning perform two standard deviations better than their peers. With modern AI, we saw an opportunity to scale this powerful, personalized tutoring model for everyone.
What We Learned & Our Innovation
We quickly realized that a simple Q&A bot wasn't enough; it can't help you with the concepts you don't know you're missing. Our true innovation is the "Depth Knowledge Tree" approach. We built an AI that doesn't just provide answers, but understands the hierarchy of knowledge. If a user is struggling, it can identify the root of the confusion and offer to explain the more fundamental, prerequisite concepts first, ensuring a solid foundation for true mastery.
How We Built It
We built Grasphoffer from the ground up on IBM's powerful AI stack. The IBM Granite model serves as the core brain for our RAG pipeline, while IBM Watson Discovery and Watson Speech to Text/Text to Speech provide the senses for document processing and voice interaction. We used a modular, agent-based architecture, starting with our core "Socratic Tutor Agent" to manage the adaptive learning conversation.
Challenges We Faced
Our greatest challenge wasn't just programming, but pedagogy. We had to teach our AI how to be a good teacher—how to detect a user's confusion and intervene at the right moment without being disruptive. Fine-tuning this human-centric logic, especially under the time constraints of a hackathon, was a difficult but rewarding process. We're proud to have built a tool that is not just smart, but wise in its approach to learning.
Log in or sign up for Devpost to join the conversation.