Inspiration

I am a Computer Science student at the University of Windsor. I wanted to build a tool that can "think" like a security expert. Most tools today look at high-level code, but I wanted to build something that could look at the raw instructions (Assembly) and find hidden dangers that others miss.

What it does

Neural-Trace is an AI assistant that checks Assembly code for security bugs. It finds "silent killers" like buffer overflows, where a hacker could crash a system or take control. It explains exactly how the bug works and even writes a new, safe version of the code in Python.

How I built it

I used Python and a tool called Streamlit to build the app. It is powered by the Gemini 3 API. I used the Deep Think feature so the AI can show its step-by-step logic. I also used Gemini's Code Execution tool to do the math and prove that a bug is real.

Challenges I ran into

The hardest part was the API "speed limits". Because the AI does a lot of deep thinking, it can be slow or hit limits quickly. I had to write special code to handle these limits so the app wouldn't crash.

Accomplishments that I'm proud of

I am very proud of the AI Reasoning Chain. It is amazing to see the AI explain exactly how a single character, like the letter 'o', can break a security flag and give a user admin access.

What I learned

I learned that AI is more than just a chatbot. When you give it the right tools, it can act like a real engineer and solve very hard technical problems.

What's next for Neural-Trace

I want to turn Neural-Trace into an extension for VS Code so programmers can check their code while they write it.

Built With

Share this project:

Updates