Skip to content
The Collection

08 · EdTech

3

input modalities

Best AI HackHackJam 2025

Learn by discovering. Guided by Rocky the Bull. An interactive AI-powered Socratic tutoring dashboard for USF featuring a real-time drawing canvas and voice input/output.

The Exhibit

01·Problem

“

Students learn best not when they are handed the answers, but when they are guided to discover the solutions themselves. Traditional AI often gives away the answer too quickly. Students need a tool that teaches through Socratic guided problem-solving across voice, text, and visual input.

02·Approach

  1. 01

    Streamlit-based multimodal interface features an interactive canvas whiteboard that allows students to draw their math problems, diagrams, or notes, which Gemini 'sees' to provide visual guidance.

  2. 02

    Google Speech-to-Text transcribes student's voice questions and Edge-TTS generates natural audio responses, making the learning experience conversational.

  3. 03

    PyPDF2 and python-docx parse uploaded course materials to give the AI context about assignments or readings.

  4. 04

    OpenRouter routes to Gemini 2.5 Pro for step-by-step Socratic tutoring that guides without revealing the answer, adapting to whiteboard drawings, uploaded files, and spoken questions.

  5. 05

    A persistent chat history tracks the conversation, allowing the AI to read the whiteboard, the uploaded files, and user queries to formulate hints and guiding questions.

03·Architecture

Swipe to explore

04·Results

3

modalities (voice, whiteboard, documents)

Socratic

guided tutoring without giving answers

Real-time

voice and whiteboard interaction

05·Stack

Streamlit · OpenRouter · Gemini 2.5 Pro · Edge-TTS · SpeechRecognition · PyPDF2 · python-docx