Building ResilAI: An AI Incident Readiness Platform with Gemini and Google Cloud

This project and article were created for the purposes of entering the #GeminiLiveAgentChallenge hackathon.

The Problem

Security teams monitor alerts constantly, but very few organizations can actually measure their readiness to respond to a cybersecurity incident.

Most tools focus on detection and monitoring rather than answering a fundamental question:

How prepared are we if an incident happens tomorrow?

The Solution: ResilAI

ResilAI is an AI-powered incident readiness platform designed to measure and communicate organizational preparedness for cybersecurity incidents.

The platform calculates deterministic readiness scores aligned with security frameworks and uses AI to translate technical findings into executive-level risk narratives.

Architecture Overview

ResilAI is built as a cloud-native platform using modern technologies:

Frontend
React + Vite

Backend
FastAPI deployed to Google Cloud Run

AI Layer
Google Gemini Flash via the Google GenAI SDK

Data Layer
Cloud SQL / SQLite (development)

Storage
Google Cloud Storage for secure report delivery

Using Gemini for Executive Intelligence

One of the key design decisions in ResilAI is separating deterministic risk scoring from AI narrative generation.

The readiness score is calculated using a rule-based engine aligned with frameworks such as:

• NIST CSF 2.0
• CIS Controls
• OWASP security guidance

Gemini is then used to translate those findings into executive-level summaries that help leadership teams understand risk exposure.

Why Google Cloud Run

Cloud Run was chosen because it provides:

• automatic scaling
• container-based deployments
• minimal infrastructure overhead

This allowed the backend API to remain stateless while scaling dynamically.

Deployment Automation

Deployment is automated using scripts that push the containerized FastAPI backend to Cloud Run.

This ensures consistent environments across development, staging, and production.

Demo

You can watch the project demo here:

https://youtu.be/Z_0aNizadoU

Repository

https://github.com/purvanshbhatt/AIRS

Final Thoughts

ResilAI demonstrates how AI can be used responsibly in cybersecurity by augmenting human decision-making rather than replacing deterministic security analysis.

The result is a system that converts complex technical findings into actionable executive intelligence.