CostPilot is a full-stack cloud analytics dashboard built to monitor deployed workloads, estimate cost usage, identify waste, and provide optimization recommendations for services running on Google Cloud Platform.
This project was built from a real problem: while deploying backend services for another project (HeliiX) on Google Cloud Run, cloud credits were being consumed faster than expected. That created the need for a lightweight internal tool that could visualize usage patterns, detect underutilized resources, and help make smarter infrastructure decisions.
While using Google Cloud Run for production workloads, I noticed a common startup/student problem:
- Credits disappear without clear visibility n- Hard to understand where cost is coming from
- No simple custom dashboard for small projects
- Existing enterprise FinOps tools are overkill for solo developers
- Need practical insights, not just raw metrics
Instead of manually checking multiple cloud consoles, I built CostPilot as a centralized optimization dashboard.
Convert cloud infrastructure metrics into understandable business decisions.
Examples:
- Low traffic at night → scale to zero n- Idle service → review necessity
- High usage → optimize concurrency
- Repeated usage spikes → investigate traffic pattern
- Professional dark-theme dashboard UI
- KPI cards (Monthly Cost, Potential Savings, Waste Score)
- Resource inventory table
- CPU utilization charts
- Savings breakdown chart
- Recommendation cards
- Auto-refresh ready architecture
- REST API endpoints
/resources/summary/recommendations- Data transformation layer
- Cloud API integration
- Google Cloud Run service discovery
- Service account authentication
- Ready for Monitoring metrics integration
- Real deployed service visibility
| Layer | Technology |
|---|---|
| Frontend | React + Vite |
| Styling | Tailwind CSS |
| Charts | Recharts |
| Backend | FastAPI |
| Language | Python |
| Cloud | Google Cloud Platform |
| Infra Service | Cloud Run |
| Auth | Service Account JSON |
You can add additional screenshots below:
costpilot/
├── frontend/
│ ├── src/
│ └── package.json
│
├── backend/
│ ├── main.py
│ ├── gcp_client.py
│ └── gcp-key.json (ignored)
│
└── README.md
Returns detected cloud services.
Returns aggregated dashboard metrics.
Returns optimization suggestions.
This project involved practical cloud engineering issues beyond coding.
The SDK was using old cached credentials instead of the intended service account key.
Fix: Explicit credential loading using Python service account auth.
Cloud APIs returned permission denied errors due to wrong project selection and disabled APIs.
Fix: Verified active project, enabled Cloud Run Admin API, corrected IAM roles.
After backend data changed, frontend cards displayed undefined values.
Fix: Recomputed metrics dynamically from /resources response.
Real metrics require correct filters, time intervals, permissions, and service labels.
Fix: Built integration incrementally rather than replacing everything at once.
It was tempting to overbuild. Instead, focus stayed on a useful MVP.
- Real cloud engineering is often configuration-heavy
- Authentication problems can consume more time than coding
- Build stable integrations first, then scale features
- Good dashboards require clear UX, not only raw data
- Solving your own real problem creates stronger products
- Real-time Monitoring API metrics
- Cost prediction using ML
- Multi-project support
- Slack / Email alerts
- PDF monthly reports
- Historical trend storage
- Role-based login system
- Kubernetes / AWS support
This project demonstrates:
- Full-stack development
- Cloud integration
- API design
- Dashboard engineering
- Cost optimization thinking
- Real-world debugging
- Product mindset
cd frontend
npm install
npm run devcd backend
pip install -r requirements.txt
uvicorn main:app --reloadCostPilot was not created as a tutorial clone. It was created to solve a real cloud cost pain point using practical engineering.
That makes it more valuable than many generic student projects.