Inspiration
Most modern homework helpers and AI tools act as answer engines—students paste a problem and immediately get the final numerical output or complete code snippet. This encourages passive copying rather than active problem-solving. We built Socratic Studio to reinvent digital tutoring by acting as an empathetic lab mentor that observes a student's screen, camera, and voice in real time, guiding them with targeted diagnostic questions without giving away answers.
What it does
Socratic Studio is a real-time, multimodal AI tutor that assists students across mathematics, physics, and coding:
- Visual Inspection: Continuously monitors a live digital whiteboard, screen share, or webcam feed showing handwritten work or code.
- Socratic Guidance: Spots step-by-step errors, formula misconceptions, or logic bugs, responding with single diagnostic prompts that lead the student to their own solution.
- Interactive UI Feedback: Triggers visual highlights, status badges, and reference formula cards on the dashboard as the user progresses.
How we built it
- Google AI Studio: Configured and tested system instructions, temperature settings, and JSON tool schema definitions.
- Gemini 2.0 Flash / Multimodal Live API: Utilized low-latency multimodal capabilities to process video frames and voice streams simultaneously.
- Web Application: Developed a web frontend with HTML, CSS, and JavaScript that hooks into Gemini via WebSockets for real-time audio and canvas feedback.
Challenges we faced
Balancing low-latency video streaming with strict prompt adherence was challenging. We had to carefully craft system instructions to ensure Gemini strictly limits itself to one Socratic question at a time and never reveals final numerical values or direct code fixes.
What we learned
We learned how to leverage Gemini's Multimodal Live API over WebSockets to build seamless, sub-second vision and audio interaction loops, as well as how to integrate structured function calling to update user interfaces dynamically.
What's next for Socratic Studio
- Adding automated subject-matter micro-quizzes generated dynamically after a student completes a problem set.
- Integrating multi-student group lab sessions where the AI tutor facilitates peer discussion.
Built With
- css3
- gemini-2.0-flash
- gemini-api
- google-ai-studio
- html5
- javascript
- websockets

Log in or sign up for Devpost to join the conversation.