Inspiration
Modern cloud networks are constantly bombarded with security threats. Monitoring tools like Dynatrace flag these issues, but security teams often struggle to analyze and patch them in real-time. We were inspired to build an intelligent, autonomous agent that not only detects threats but instantly tells you how to fix them using AI.
What it does
CyberShield AI Agent is a next-generation security dashboard. It integrates cloud observability data with the Google Gemini API. The dashboard features a dual-mode system: a Demo Mode to simulate active threats (like SQL Injection or Log4j) and generate comprehensive AI Analysis Reports with step-by-step mitigation plans, and a Live Mode that confirms a 100% secure infrastructure status.
How we built it
We developed the core logic using Python and built the user interface with Streamlit, customizing it with a premium cyber-dark and neon-glow CSS theme. We utilized the Google Gemini API as our cognitive AI engine to ingest simulated threat logs, evaluate overall system safety percentages, and output actionable security patches. Version control and documentation were managed entirely on GitHub.
Challenges we ran into
One of the main challenges was designing a dynamic UI that could seamlessly switch between live secure telemetry and high-threat demo logs without crashing. Fine-tuning the prompt engineering for Gemini to ensure it consistently returned structured, accurate mitigation steps rather than generic advice was also a rigorous process.
Accomplishments that we're proud of
We are incredibly proud of creating a sleek, highly responsive, and production-ready security dashboard. Achieving an end-to-end simulation where the AI cuts down incident response times from hours to seconds—delivering an instant threat analysis report—is a huge win for us.
What we learned
We gained deep insights into combining enterprise observability tools with Generative AI. We mastered custom theme engineering in Streamlit, advanced prompt structuring for security contexts using the Gemini API, and learned how to model real-world vulnerabilities like RCE and SQL injections realistically.
What's next for CyberShield AI Agent
The next step is to fully productionize the live data pipeline by directly integrating live webhooks from a Dynatrace tenant. We also plan to implement an automated patching feature where the AI agent can safely execute scripts to fix minor vulnerabilities autonomously with human-in-the-loop approval.
Log in or sign up for Devpost to join the conversation.