In a webinar hosted by The Top Voices, MLOps Engineer Ivan Timonov shared a practical approach to managing cloud costs on Google Cloud. The session focused on adopting FinOps principles and reducing spending without heavy tools, complex platforms, or long implementation cycles.
Speaker
Ivan Timonov is an MLOps Engineer at Tabby, a UAE-based BNPL fintech unicorn. He works on optimizing cloud infrastructure, developing FinOps practices, and building systems that help teams gain better visibility and control over cloud costs.
The Challenge of Unpredictable Costs
Startups running on GCP often experience sudden billing spikes. Common drivers include forgotten GPU instances, oversized VMs, inefficient BigQuery queries, excessive log retention, and unattached disks or snapshots. Without dedicated FinOps teams, cost management tends to be reactive, making budgets difficult to control and forecasts unreliable.
A Three-Step Framework
The approach is built around three core principles: Visibility → Standards → Automation.
Visibility. The first step is establishing a clear picture of spending. Billing Reports and Cost Breakdown help identify where money goes and who owns it. Mandatory labels ensure resource attribution, while budgets and alerts at 50%, 80%, and 100% thresholds enable early detection of overspending. Exporting billing data to BigQuery and creating dashboards in Looker Studio makes tracking trends and ownership simple and visual.
Standards. Once visibility is in place, lightweight rules help prevent most cost leaks. These include mandatory resource labeling, automatic shutdown of non-production environments, BigQuery optimization through partitioning and clustering, lifecycle policies for Cloud Storage, and switching to slot commitments for stable workloads where predictable pricing brings significant savings.
Automation. The final step is embedding best practices into tools and workflows. Cloud Scheduler automates resource shutdowns, while Recommender API delivers optimization insights. Automated cleanup scripts handle unused disks, snapshots, and temporary buckets. Terraform policies ensure resources without required labels cannot be created, reducing manual review overhead.
Results and Key Metrics
The goal is not only reducing spending but making it predictable. Business KPIs include keeping monthly costs within ±15% of budget and reducing unattributed spending close to zero. Engineering metrics focus on label coverage, BigQuery optimization, lifecycle policy adoption in GCS, and reaction time to budget alerts.
Shifting from BigQuery’s on-demand pricing to slot commitments, combined with partitioning and query optimization, can significantly lower expenses while stabilizing billing. In some cases, savings can reach $100,000 per month.
Conclusion
Effective cloud cost control on GCP doesn’t require complex FinOps platforms or large teams. By focusing on visibility, lightweight standards, and automation, startups can reduce unexpected bills, improve budget predictability, and make faster, more confident product decisions.