Cloud Cost Optimizer

The Problem

A startup's GCP bill had been growing 15% month-over-month for six months, with no clear explanation. The engineering team had no visibility into which resources were actually being used versus idle. Oversized VM instances from a load test three months ago were still running. Reserved IPs sat unattached. Persistent disks remained after clusters were deleted.

Nobody had time to audit infrastructure manually. Every sprint was focused on feature delivery, and the cloud bill was treated as an invisible cost — until it crossed $30K/month and finance raised the alarm.

The Solution

I built a two-layer system: a Python scanning layer that interrogates the GCP API in depth, and an n8n orchestration layer that schedules scans, processes reports, and handles the approval workflow for any changes. Every morning, the system produces a prioritized list of optimization opportunities. Safe changes — like releasing unattached IPs or deleting empty disks — are applied automatically. Anything that could affect running workloads goes through a Slack approval gate before execution.

• Python scripts scan all GCP projects for idle VMs, oversized instances, unattached disks, and orphaned IPs
• A cost analyzer calculates the monthly spend of each wasteful resource and ranks by savings potential
• n8n generates a daily optimization report and posts a summary to a Slack channel
• Safe cleanup actions are applied automatically without human intervention
• Impactful changes (VM downsizing, instance termination) require a Slack button approval before execution
• Weekly summary reports track total savings and trend over time

Pipeline

The Results

Within 30 days of deployment, monthly GCP spend dropped from $30K to $18K — a 40% reduction. The bulk of savings came from right-sizing oversized instances and cleaning up forgotten resources that had accumulated over 18 months of rapid growth.

• $12K/month in direct savings, sustained for 6+ months since launch
• 40% reduction in total cloud spend without any impact on production performance
• Auto-scaling became viable once instances were right-sized — the team had avoided it before due to unpredictable costs
• Engineering team spends zero time on manual infrastructure audits
• Finance now has weekly cost trend reports delivered automatically to their inbox