🚀 Scorix AI — AI Evaluation & Ranking Platform

Production-grade AI evaluation system that scores, ranks, and improves AI-generated responses using machine learning and real-time APIs.

---

🌟 Overview

Scorix AI is a full-stack AI benchmarking platform that simulates enterprise-level LLM evaluation systems.

It allows you to:

• ⚡ Evaluate AI responses using ML models
• 🧠 Rank multiple responses intelligently
• 📊 Track evaluation logs & feedback
• 🔁 Retrain models with new datasets

---

🎯 Key Features

🧪 AI Evaluation Engine

• Scores responses using:

  • TF-IDF vectorization
  • Cosine similarity
  • Feature engineering

• Output: 0–10 quality score

---

🏆 Ranking System

• Compare multiple responses for a prompt
• Returns ranked list based on score

---

📊 Feedback Learning (RLHF-style)

• Store human feedback
• Use for retraining and improvement

---

🔁 Dataset Upload + Retraining

• Upload CSV dataset
• Retrain model via API

---

🧠 ML Model

⚙️ Architecture

TF-IDF (5000 features)
+ Cosine Similarity
+ Length Features
+ Word Count
+ Keyword Overlap
→ Gradient Boosting Regressor

---

📈 Model Performance

• 🔥 R² Score: 0.9658
• 🔥 MAE: 0.2558
• 📊 Dataset Size: 50,000 samples

---

🧠 Architecture

Frontend (Vanilla JS)
        ↓
FastAPI Backend
        ↓
ML Model (TF-IDF + Gradient Boosting)
        ↓
Evaluation + Ranking Engine
        ↓
Database (SQLite)

---

🛠️ Tech Stack

Layer	Technology
Backend	FastAPI
ML Model	Scikit-learn
NLP	TF-IDF + Cosine Similarity
Database	SQLite
ORM	SQLAlchemy
Frontend	HTML, CSS, Vanilla JS
Server	Uvicorn

---

⚡ Quick Start

1️⃣ Install dependencies

pip install -r requirements.txt

---

2️⃣ Run server

python -m uvicorn backend.main:app --reload

---

🌐 Open

• App → http://localhost:8000
• Docs → http://localhost:8000/docs

---

📡 API Endpoints

🔹 Evaluate Response

POST /evaluate

{
  "prompt": "What is AI?",
  "response": "AI is the simulation of human intelligence in machines."
}

---

🔹 Rank Responses

POST /rank

{
  "prompt": "What is AI?",
  "responses": [
    "AI is computer intelligence.",
    "AI is the simulation of human intelligence in machines.",
    "AI is random."
  ]
}

---

🔹 Submit Feedback

POST /feedback

---

🔹 Upload Dataset + Retrain

POST /upload-dataset

---

🔹 Health Check

GET /health

---

📸 Screenshots

---

🧠 What Makes This Special

✔ End-to-end ML pipeline (dataset → training → deployment) ✔ Real-time scoring & ranking APIs ✔ Feature-engineered ML model (not just black-box) ✔ RLHF-style feedback system

---

🚀 Future Improvements

• 🔁 Reinforcement learning (RLHF loop)
• 📊 Model comparison leaderboard
• ⚡ Async batch evaluation
• 🧠 Deep learning-based scoring

---

👨‍💻 Author

Shivansh Thakur Linkedin

---

⭐ Support

If you like this project, give it a ⭐ on GitHub 🚀