Скачать или смотреть Building SentriGuard AI – A Modular AI Safety Gateway (Hackathon Submission)

Welcome to the official demo of SentriGuard AI, our project for the CODE RED Hackathon.

SentriGuard AI is a real-time safety gateway designed to protect Large Language Models (LLMs) from prompt injection attacks, policy overrides, and malicious user inputs.
As AI becomes more powerful and widely used, securing the input layer is essential—and that’s exactly what our system is built to do.

🔐 What SentriGuard AI Does

Detects harmful or malicious prompts before they reach the LLM

Blocks common jailbreak patterns like “Ignore previous instructions”

Cleans unsafe requests with a smart sanitization layer

Uses ML + embeddings to identify hidden or complex prompt attacks

Provides real-time risk scoring and decision-making (Allow / Block / Sanitize)

⚙️ Key Features

Multi-layer defense pipeline:
Rule Engine → Sanitizer → ML Classifier → Embedding Detector → Decision Layer

Lightweight and fast (no GPU required)

API-driven, modular, and easy to integrate

Clean dashboard for visibility and debugging

Built for enterprise-grade AI applications

🎥 What’s inside this video

The problem: Why LLMs are vulnerable

Live demo with safe vs malicious prompt tests

Architecture and pipeline walkthrough

Technical explanation of how SentriGuard AI works

Future roadmap and real-world applications

🙌 Team Behind SentriGuard AI

Team Name: ProtoLab
Hackathon: CODE RED 3.0