Pero AI

🔐 PeroAI – Multi-Agent GenAI Firewall on AWS

PeroAI is a multi-agent governance system that detects, explains, rewrites, and audits user prompts to ensure enterprise safety and compliance in GenAI applications.

Built for the AWS + Impetus GenAI Hackathon, PeroAI tackles the growing need for trust and governance in LLM usage by combining smart UX and a Lambda-based architecture.

💡 Inspiration

As GenAI becomes embedded in enterprise workflows, it introduces risk: jailbreaks, privacy leaks, unsafe prompts. Enterprises need guardrails—not just LLMs.

PeroAI was inspired by this problem. It aims to be a smart firewall between users and AI, enforcing safe, ethical, and compliant use of LLMs like Claude and GPT.

⚙️ What It Does

✅ Detects harmful or sensitive content in prompts
🧠 Explains policy violations in plain language
✍️ Rewrites unsafe prompts using Amazon Bedrock (Claude v2)
🗂 Logs every interaction for audit/compliance in MCP format
🧩 Runs on AWS Lambda with optional Step Function orchestration

🛠 How We Built It

✅ Frontend

Next.js + React UI to input prompts and show results
Ready for API integration with Bedrock + Lambda
Clean UX with optional 3D avatar interface

🚧 Backend (In Progress)

Designed multi-agent system with four Lambda agents:
- GuardAgent – detects violations
- PolicyAgent – explains them
- RewriteAgent – rewrites using Claude
- AuditAgent – logs using MCP
Uses Amazon Bedrock, DynamoDB, and AWS Lambda