Our Landing Page
Our Browser Extention
Our Kubernet cluster for deploying Kali Linux VMs
https://youtu.be/fPG_FeuLvRE?si=XpYcpKt7QY50dyXi

My Journey Building AI CyberShield

🌟 Inspiration

One day, while sitting with a friend, I heard about the endless headache of browsing multiple websites and testing different cybersecurity tools. The process took hours. That sparked a thought:

$$ \text{What if AI could automate vulnerability testing?} $$

When I dug deeper, I realized something even more important: cybersecurity is a privilege. Large companies can afford teams of security experts, but startups and small-to-medium businesses (SMBs) often cannot. This became my motivation — to create a system so simple that even a non-technical founder could find vulnerabilities in their applications.

📚 What I Learned

This project pushed me out of my comfort zone and into new territories. I learned:

Kubernetes for orchestration and scaling.
Agent orchestration for coordinating multiple AI-driven tasks.
Kali Linux commands for real-world penetration testing.
Docker for containerizing agents and services.
How to secure command execution so agents don’t harm the host system.
Designing a browser extension to extract APIs and authentication tokens hidden behind login walls.

🛠️ How I Built It

I worked solo on this project, tackling different parts in parallel:

Frontend & Site – Built first using Next.js and Prisma with MongoDB for storage.
Browser Extension – To capture session data, APIs, and tokens behind authentication walls.
Agentic System – AI agents orchestrated using Python, FastAPI, and OpenAI models, with Cloudflare AutoRAG for context retrieval.
Infrastructure – Deployed using Kubernetes and Docker on Azure.
Remote Execution Layer – Agents execute Kali Linux commands in isolated VMs. If something goes wrong, Kubernetes can scale down the compromised VM and spin up a fresh one.

⚔️ Challenges I Faced

Like any meaningful project, this one wasn’t easy. Some of the key challenges were:

Scalable Deployment in Kubernetes – Ensuring the system could handle multiple concurrent scans.
Security Risks in Kali Linux VM – Preventing harmful tool executions from breaking the system.
Agent Timeout & Orchestration – Managing AI agents with different roles, iteration limits, and error handling.
Learning Curve – I had to teach myself new technologies quickly and apply them immediately.

I overcame these challenges by separating concerns: agents now run commands remotely in isolated VMs. If anything breaks, the system can self-heal using Kubernetes scaling.