vdda

Inspiration

Traditional competitive exam preparation heavily relies on static multiple-choice questions (MCQs) and rote memorization. However, modern competitive exams have evolved toward complex analytical evaluation—featuring statement-based questions, assertion-reasoning, and multi-column matching. We created vdda to bridge the gap between outdated study tools and the real, analytical demands of modern competitive exams, giving students a realistic simulation environment powered by AI.

What it does

vdda is an intelligent exam readiness and simulation platform that transforms how students prepare for competitive tests:

• Advanced Question Formats: Simulates modern testing patterns including assertion-reasoning, multi-statement evaluation, and matching columns.
• Multimodal Note Evaluation: Evaluates handwritten notes directly from student uploads, offering instant corrective feedback and gap identification.
• Topic-Level Progress Tracking: Pinpoints micro-concept strengths and weaknesses to direct student effort where it matters most.
• Timed Mock Simulations: Provides realistic, time-bound exam conditions with detailed post-test analytical breakdowns.

How we built it

• Frontend: Built a responsive, interactive user interface using React Native / Flutter for cross-platform availability.
• Backend: Developed high-performance REST APIs using Python FastAPI / Node.js to manage test states and user data.
• AI Engine & Vision: Integrated vision-capable Large Language Models (LLMs) paired with Retrieval-Augmented Generation (RAG) to accurately digitize, evaluate, and provide feedback on handwritten notes against syllabus standards.
• Database & Vector Search: Utilized PostgreSQL for user profiles and test metadata alongside Vector Databases for embedding-based concept matching.

Challenges we ran into

• Handwriting OCR Accuracy: Processing varying handwriting styles, messy layouts, and diagrammatic notes required fine-tuning computer vision prompts to avoid false corrections.
• Domain Accuracy in AI Feedback: Ensuring AI-generated explanations for complex assertion-reasoning questions remained 100% grounded without model hallucinations.
• Real-time Latency: Minimizing latency during live, timed mock simulations while processing complex analytical questions and instant score updates.

Accomplishments that we're proud of

• Successfully building an end-to-end multimodal pipeline that accurately parses and grades handwritten notes.
• Designing a seamless simulation UI that replicates modern competitive exam environments.
• Creating a topic-level granular tracking system that gives students actionable insights instead of simple percentage scores.

What we learned

• The crucial role of grounding LLMs through RAG to ensure absolute factual accuracy for exam prep materials.
• Strategies for optimizing computer vision workflows for student-submitted media.
• Design patterns for presenting dense analytics and progress metrics in a simple, non-overwhelming student dashboard.

What's next for vdda

• Multilingual Evaluation: Expanding OCR and AI feedback to support regional Indian languages (e.g., Kannada, Hindi).
• AI Adaptive Timelines: Automatically generating daily study plans based on mock simulation weaknesses.
• Offline Mock Mode: Enabling students in low-bandwidth areas to download test packages and sync performance metrics once reconnected.

What's next for vdda

Built With

• gemini