Summary
BrokerEdge AI is an end-to-end real estate SaaS platform that helps brokers recruit, retain, and scale their operations with a CRM, MLS sync, onboarding workflows, and a built-in AI Copilot. My DevOps role was to architect, automate, and optimize the infrastructure for scalability, cost-efficiency, and secure AI/MLS integration.
Gallery
Objectives
- Automate deployments and environment configs
- Secure AI and MLS data pipelines
- Optimize Firestore rules and usage
- Enable scalable AI interactions using Genkit
- Build a real-time dashboard with monitoring hooks
Technical Challenges & Solutions
Problem 1: Manual Deployments Slowed Down the Team
"Deployments were initially manual via Firebase CLI and Vercel UI. This delayed testing, introduced human error, and lacked rollback visibility."
Solution: CI/CD with GitHub Actions + Vercel
.github/workflows/deploy.yml
name: CI/CD Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- run: npm install
- run: npm run lint && npm run build
- name: Deploy to Vercel
uses: amondnet/vercel-action@v20
with:
vercel-token: ${{ secrets.VERCEL_TOKEN }}
vercel-org-id: ${{ secrets.VERCEL_ORG_ID }}
vercel-project-id: ${{ secrets.VERCEL_PROJECT_ID }}
scope: personalResults:
- Deployment time reduced from ~10 min to ~1.2 min
- Every push to main auto-deployed to staging or production
- Built-in validation ensured cleaner, safer code delivery
Problem 2: AI Copilot Was Too Slow on Cold Starts
"Firebase Cloud Functions were experiencing cold start lag on AI generation requests, impacting the Copilot experience."
Solution: Pre-warm AI Copilot Function & Queue System
functions/genkit.ts
import { initializeGenkit } from '@genkit-ai/core';
import { onCall } from 'firebase-functions/v2/https';
initializeGenkit();
export const copilotChat = onCall(async (req) => {
const { input, userContext } = req.data;
const result = await myGenkitFlow({ input, userContext });
return { response: result.output };
});Results:
- Added regional warm-up ping every 5 minutes via a scheduled pingCopilot() job
- Reduced AI response latency to <1.2s avg
- Implemented retry queue for failed requests using Firestore batching
Problem 3: Firestore Costs Were Climbing with Agent Pipelines
"Thousands of Firestore reads were triggered by poorly scoped client queries."
Solution: Query Optimization + Denormalized Structure
firestore-query.js
// BEFORE (inefficient)
const leads = await getDocs(collection(db, 'clients'));
// AFTER (scoped & indexed)
const q = query(
collection(db, `clients/${{userId}}/leads`),
where('status', '==', 'active'),
orderBy('createdAt', 'desc'),
limit(50)
);
const snapshot = await getDocs(q);Results:
- Firestore reads reduced by 72%
- Monthly cost dropped from ~$180 → ~$45
- Faster rendering of client pipelines in dashboard
Problem 4: Lack of Role-Based Security & Data Isolation
"Without well-defined access rules, junior agents could accidentally access brokerage-wide documents."
Solution: Firebase Security Rules + Claims
firestore.rules
match /clients/{agentId}/{clientId} {
allow read, write: if request.auth.uid == agentId;
}
match /adminData/{doc} {
allow read: if request.auth.token.role == "admin";
}Results:
- Implemented broker/admin/agent roles via Firebase Auth custom claims
- Isolated views and write access per agent/team
- Passed internal security audit
DevOps Stack
Firebase
Hosting, Firestore, Auth, Functions
Genkit AI
AI Copilot orchestration
GitHub Actions
CI/CD, deployments
Vercel
Hosting frontend + dashboard
Docker
Dev containers for MLS services
Google Cloud IAM
Security + role control
Cloud Scheduler
Ping + AI warm-up jobs
Sentry + Logs
Error tracking & monitoring
Key Outcomes
- CI/CD enabled faster testing and zero-downtime deployments
- AI Copilot response latency dropped to ~1.2s
- Firebase secured with role-based rules & scoped Firestore queries
- Reduced Firestore billing by 70% through optimized reads
- Ready to scale from 10 → 10,000+ users on Firebase+Genkit
Reflection
Working on BrokerEdge AI sharpened my ability to balance DevOps automation, security, and AI orchestration in a fast-moving SaaS environment. This project is a great example of delivering speed, stability, and smart infrastructure at scale.