cyberaware-ai

• 

  result 2

• 

  total result of all scenarios

• 

  last page

• 

  Starting page

• 

  result

• 

  scenario generated by nemotron

Inspiration

Most security training is the same boring quiz for everyone. But real scammers go after your specific weak spots. So I wanted a trainer that figures out what you fall for and practises that with you.

What it does

CyberAware AI throws realistic fake phishing, scam and social-engineering messages at you. You pick what to do, and NVIDIA Nemotron tells you if it was safe, risky or dangerous, plus which red flags you caught and which you missed.

Then it works out your weak spot (like urgency tricks or fake authority) and writes the next scenario to test exactly that. Fall for urgency tricks? You get more of them, in new situations. Start nailing them? It gets harder and moves on to something new.

After 5 rounds you get a profile with your scores, your biggest weakness, your strength, and a round-by-round look at how the training adapted.

How we built it

• Next.js, React, TypeScript and Tailwind, hosted on Vercel
• Nemotron (nvidia/nemotron-3-super-120b-a12b) running on Nebius Token Factory
• Nemotron writes every scenario and judges every decision
• A small adaptive engine tracks your weak spots and picks what comes next
• If the AI gives a messy answer, the app checks it, retries, or falls back to a practice scenario
• The API key stays on the server, and every attack in the app is fictional

Challenges we ran into

Getting clean, reliable answers out of an AI model, keeping the app working when the AI is slow, and making it actually adapt instead of just being a random quiz.

Accomplishments that we're proud of

It's a full working app where the AI does real work every round, and the training really does change based on your mistakes.

What we learned

How to build and deploy a full AI app, how to use Nebius Token Factory, and how to check AI output before showing it to users.

What's next for cyberaware-ai

An AI-written final recommendation, saved progress, and more attack types.

Built With

• github
• nebiustokenfactory
• next.js
• nvidianemotron
• react
• tailwindcss
• typescript
• vercel