🚀 VeridicAI AI-Powered Coding Assignment Evaluator
Fair. Fast. Intelligent.
🌍 Live Deployment
Frontend (Vercel): https://veridicai.vercel.app
Backend API (Render): https://your-backend-url.onrender.com
Swagger Docs: https://your-backend-url.onrender.com/docs
📌 Overview
VeridicAI is an AI-powered coding assignment evaluator that goes beyond traditional test-case validation. It evaluates submissions across multiple dimensions — correctness, efficiency, readability, structural quality, and edge-case handling — delivering structured, human-like feedback in under two minutes.
Designed for hiring platforms, educators, bootcamps, and competitive programming environments, VeridicAI provides scalable, explainable, and fair automated evaluation.
🎯 Problem Statement
Traditional code evaluation systems face several challenges:
⏳ Delayed or manual grading
📉 Binary pass/fail test case scoring
❓ Lack of structured feedback
⚖️ Inconsistent evaluation standards
📈 Poor scalability for large submissions
VeridicAI addresses these limitations through a multi-dimensional automated evaluation engine that provides consistent, explainable, and performance-aware scoring.
🧠 Core Features ✅ 1. Correctness Evaluation
Executes submitted code against predefined test cases
Calculates pass/fail ratio
Detects incorrect edge-case handling
Generates correctness score
⚡ 2. Efficiency Analysis
Cyclomatic complexity detection using Radon
Identifies nested logic structures
Encourages optimal algorithmic patterns
Provides performance classification
📖 3. Readability & Code Quality
Static analysis using Pylint
Evaluates naming conventions
Checks structural clarity
Encourages maintainable code practices
🗣️ 4. AI-Generated Feedback
Structured, human-like improvement suggestions
Clear explanation of weaknesses
Actionable recommendations
🌐 5. Professional Dashboard UI
Clean SaaS-style interface
Real-time evaluation display
Performance visualization
Fully responsive design
📊 Scoring Model Dimension Weight Correctness 50% Efficiency 30% Readability 20% Final Score Formula Overall Score = 0.5 × Correctness
- 0.3 × Efficiency
- 0.2 × Readability
This ensures balanced evaluation beyond surface-level validation.
🏗️ System Architecture Frontend (React + TailwindCSS) ↓ FastAPI REST API ↓ Evaluation Engine ├── Test Case Runner ├── Complexity Analyzer (Radon) ├── Readability Analyzer (Pylint) └── Feedback Generator
🔄 How Evaluation Works
User uploads Python file or GitHub RAW link
Backend securely executes code in isolated environment
Test cases are run using subprocess
Complexity analysis is performed using Radon
Readability analysis is performed using Pylint
Weighted score is calculated
Structured feedback is generated
Results are returned to frontend dashboard
Total response time: ~1–2 seconds (local) / ~2–5 seconds (production).
🛠️ Tech Stack Frontend
React (Vite)
TailwindCSS
Custom CSS animations
Backend
FastAPI
Uvicorn
Radon
Pylint
Requests
Deployment
Vercel (Frontend)
Render (Backend)
🚀 Running Locally Backend Setup pip install -r requirements.txt uvicorn main:app --reload
Backend runs at:
Frontend Setup npm install npm run dev
Frontend runs at:
📤 Supported Input Methods
Upload Python file
Provide GitHub RAW file URL
📈 Sample Output Overall Score: 87/100
Correctness: 100/100 Efficiency: 90/100 Readability: 50/100
Feedback:
- All test cases passed.
- Efficient implementation detected.
- Improve variable naming and documentation.
📊 Evaluation Dimensions
✔ Correctness ✔ Edge-case handling ✔ Algorithmic efficiency ✔ Code maintainability ✔ Structural clarity
🔐 Security & Isolation
Code execution handled via subprocess with timeout
Temporary file storage with automatic cleanup
Execution timeout to prevent infinite loops
Controlled evaluation environment
📈 Scalability Potential
VeridicAI can be extended to support:
Multi-language evaluation (C/C++/Java)
Plagiarism detection
Submission history tracking
Role-based evaluation
Leaderboard and benchmarking
Cloud database integration
AI-powered optimal solution comparison
👨💻 Author
Veeraj Ratrikar
💡 Vision
To redefine automated code evaluation by combining fairness, performance intelligence, and AI-driven feedback into one scalable and explainable platform.