GUARDIAN AI

• 

  DASHBOARD OVERVIEW

• 

  LIVE EVENTS

Inspiration We noticed companies racing to deploy AI chatbots and assistants without any visibility into security threats, costs, or quality—flying blind with million-dollar stakes.

What it does

GuardianAI monitors every LLM request in real-time, using Gemini AI to detect 6 threat types: prompt injections, PII leaks, jailbreaks, toxic content, cost spikes, and quality degradation.

How we built it

Challenges we ran into

Vertex AI required Terms of Service acceptance we couldn't complete, so we pivoted to Google AI Studio API (same Gemini models) while maintaining the architecture's integrity.

Accomplishments that we're proud of

What we learned

What's next for GUARDIAN AI

Expand to multi-LLM support with cost comparison (GPT-4 vs Claude vs Gemini), A/B testing for prompts, and compliance reporting (GDPR, HIPAA audit trails)—becoming the Datadog for the AI era.

Built With

• css
• datadog
• firebase
• python
• react
• typescript
• vertex