NeuroVault is a full-stack system designed to store, manage, and analyze AI training processes using a structured relational database. It extends traditional database applications by incorporating real-time analytics, anomaly detection, and a web-based visualization layer.
- Design a normalized relational database for AI training data
- Implement database connectivity using a backend service
- Develop a front-end interface for user interaction
- Perform CRUD operations through a web application
- Provide analytical insights from stored data
Frontend (HTML, Tailwind CSS) → Backend (Flask, Python) → PostgreSQL Database
Optional: Raspberry Pi for distributed data ingestion
- PostgreSQL (Relational Database)
- Flask (Backend Framework)
- psycopg2 (Database Connectivity)
- HTML + Tailwind CSS (Frontend)
- Chart.js (Data Visualization)
The system is based on a normalized schema (BCNF), including the following entities:
- MODEL
- MODEL_VERSION
- TRAINING_RUN
- EXECUTION_STEP
- METRIC_SCALAR
- DATASET
- RUN_DATASET_MAP
Additional tables:
- ALERT
- SYSTEM_LOG
- Insert new training metrics
- Retrieve and display data
- Update metric values
- Delete records
- Line chart for loss trends
- Tabular representation of metrics
- REST-style routing using Flask
- Secure parameterized SQL queries
- Trigger-based alerts for abnormal values
pip install flask psycopg2
- Create database:
CREATE DATABASE aimemory;
- Execute SQL scripts from previous phases
python app.py
aimemory_da3/ │ ├── app.py ├── templates/ │ ├── index.html │ ├── insert.html │ ├── update.html │ ├── static/ └── requirements.txt
- Real-time streaming using WebSockets
- Advanced anomaly detection using machine learning
- Authentication and user management
- Cloud deployment
C. Kumaran B.Tech Computer Science (Data Science)