Inspiration
Across the UK and globally, millions of children from disadvantaged backgrounds face persistent academic attainment gaps. In England alone, only 25% of disadvantaged pupils achieved grade 5 or above in English and Maths GCSEs in 2022/23, compared to 52% of their peers. Barriers include limited access to quality tutoring, lack of tailored support, and insufficient parental engagement.
Key Insight: Underprivileged students need accessible, personalized, and engaging learning experiences that bridge academic gaps and empower families.
What it does
My platform delivers a multimodal tutor experience that adapts to each learner’s needs, validates content for safety and accuracy, and actively involves parents. By combining Gemini 3’s state-of-the-art reasoning and multimodal capabilities via Gemini 3, we democratize access to high-quality tutoring.
Platform Benefits TutorBuddy is a multimodal platform that transforms a seed topic or a photo of a textbook page into a comprehensive Mastery Canvas. It moves beyond "chat" to create a structured pedagogical experience.
✨ Platform Benefits • Personalized Learning: Gemini 3 adapts explanations and quizzes to individual student profiles in real-time. • Multimodal Accessibility: Supports text, scanned textbook pages, and custom video explainers to lower barriers for students with diverse needs. • Validated Content: A dedicated Validator Agent ensures every word is accurate, age-appropriate, and safe. • Parental Engagement: Automatically generates reports for guardians, fostering family involvement and compliance. • Cost Efficiency: Built on a cloud-native architecture (Vercel Fluid Compute) for rapid scaling.
The system supports seed topics and scanned textbook uploads, generating interactive learning bytes that include: • Fun, engaging 101 explanations with tips and weird facts • Short, custom video explainers • Bloom’s Taxonomy-aligned quizzes for assessment and mastery
Impact Metrics: We aim to increase engagement, improve assessment scores, and foster parental involvement, with measurable outcomes such as quiz pass rates, time-to-competence, and parental feedback
How we built it
• User Interface (Vercel v0): Accepts seed topic, scanned textbook pages, and user preferences.
• Main Tutor Orchestrator (Gemini 3 Pro): Processes inputs, extracts key concepts, generates explanation, video prompt, and quiz.
• Validator Agent (Gemini 3 Flash): Receives generated content, runs parallel validation checks (accuracy, safety, bias), updates session state.
• Parent Reporter Agent (Gemini 3): Summarizes validated session, generates parent report, updates learner profile.
• State Management Layer: Persistent learner state (mastery, misconceptions, engagement signals), session logs, privacy controls.
• External Services: OCR (Google Vision, Tesseract), Video Generation (Veo 3.1), Data Storage (Firebase/Firestore), Analytics.
✨ Key Differentiators (The "Wow" Factors) Agentic Orchestration (Gemini 3 Flash)
Instead of a simple chat, TutorBuddy acts as an agent that plans a curriculum. It validates safety, researches the topic using Google Search Grounding, and structures a pedagogical lesson plan in real-time. Feature: Contextual OCR. Students can snap a photo of their homework, and TutorBuddy aligns its lesson with their specific school curriculum. Hyper-Personalized Multi-Modality
Visual Learners: Receive 5 custom AI-generated diagrams and illustrations (Gemini 2.5 Image). Auditory Learners: Listen to an emotional, multi-speaker podcast-style dialogue between "Buddy" (the tutor) and "Sam" (the curious student) using Gemini 2.5 Native Audio. Kinesthetic/Active Learners: Interactive SVGs and "Deep Dive" micro-lessons that allow exploring tangents instantly. Guardian Insights Loop
While the child learns, TutorBuddy acts as a consultant for the parent. Generates a dedicated Parent Report summarizing what was learned, highlighting cognitive wins, and suggesting offline activities to reinforce learning. Safety-First Architecture
Built-in "Pre-Check" agent that screens topics for age-appropriateness and safety before generation begins.
Challenges we ran into
The Pivot: We moved away from a purely agentic approach to a multimodal-first, safe learning experience. We learned there is no need to force agentic capability unless it truly serves the user.
Accomplishments that we're proud of
Rapid delivery of a robust, base-level platform that feels market-ready. Creating a tool that balances high-tech AI reasoning with real-world educational safety. Successfully integrating Bloom’s Taxonomy-aligned quizzes for true mastery assessment.
What we learned
Pivoted away from agentic to non-agentic but with a greater focus on multi modal , safe learning experience There is no need to force agentic capability unless needed
What's next for tutorbuddy -
Continue to build and release an improved MVP to the market • Pilot Programs: Deploy in partner schools and community organizations • Open Source: Release core agentic workflows for adaptation and scaling • Sponsorships and Grants: Seek funding from educational charities, government innovation funds, and corporate sponsors
🎯 Hackathon Judging Checklist [x] Social Impact: Directly addresses the UK attainment gap for disadvantaged pupils. [x] Technical Complexity: Utilizes Gemini 3 Pro, Flash, and Veo 3.1 in a unified pipeline. [x] Innovation: Uses latest Gemini models for a validated, safe multimodal workflow. [x] Design: Polished, child-friendly, and accessible UI.
Reasoning Engine -> Gemini 3 Flash Selected for its sub-second latency and complex instruction following. It handles the "Orchestrator" role, managing 5 parallel streams without hallucinating safety rules. Visuals Gemini 2.5 Flash Image -> Generates consistent, educational style illustrations instantly, crucial for visual learners. Audio/TTS Gemini 2.5 Flash Audio -> The Multi-Speaker capability allows us to simulate a Socratic Seminar (Student asking, Teacher answering) rather than a mono-tone lecture. Diagrams-> Gemini 3 Flash (Code) Used to generate raw SVG code. Gemini 3's coding ability allows it to draw perfect vector diagrams of cells, atoms, or geometry. Frontend -> React 19 + Tailwind High-performance, responsive UI with fluid animations.
Built With
- gemini3
- nextjs
Log in or sign up for Devpost to join the conversation.