
Application Security requires massive compute to analyze raw HTTP traffic intelligently. I wanted to build a tool that uses the reasoning capabilities of Llama 3.1 70B to hunt for vulnerabilities like IDOR, SSRF, and Path Traversal, but I didn't want to pay for a 24/7 cloud GPU. AppSec Hunter automates the intelligence extraction and then instantly destroys the backend infrastructure to freeze billing. I built a custom Python API bridge that pipes raw intercepted network traffic from a local environment directly into Llama 3.1 70B running on an enterprise AMD MI300X cloud instance. The AI acts as a programmatic scanner, structuring its findings into strict JSON. To optimize cloud economics, I designed a "Ghost Backend" architecture: the exact second the intelligence is extracted, the cloud instance is destroyed. The data is then visualized on a zero-cost Streamlit dashboard, which I deployed to Hugging Face Spaces for community access.
10 May 2026