"Financial discipline isn't the opposite of customer-centricity. It's what enables it."
Most organizations treat infrastructure optimization and customer experience as competing priorities.
Cut costs → less innovation. Invest in CX → higher cloud bills. Pick one.
That assumption is wrong. And expensive.
This project proves it.
Over 24 months, a multi-cloud infrastructure optimization program delivered $5.4M in savings — 59% cost reduction — while simultaneously funding AI-powered customer features, faster experiences, and better personalization.
Zero SLA degradation. Zero customer impact. Zero compromise.
The infrastructure got leaner. The customer experience got better. At the same time.
| Metric | Result |
|---|---|
| Total savings over 24 months | $5.4M (59% reduction) |
| AI/ML workload savings | $1M+ annually |
| Sustained efficiency across departments | 41% |
| Budget utilization consistency | 39.1% – 43.3% |
| Departments under budget | All 8 — for 24 consecutive months |
| SLA degradation during optimization | Zero |
| Savings reinvested in CX innovation | 100% |
Why AI/ML Infrastructure Is the Hidden FinOps Opportunity
AI workloads now represent 30%+ of enterprise cloud spend.
Most organizations are flying blind on this cost center — paying full price for compute that runs at 40% utilization, training models on reserved instances that should be on spot, and running inference on infrastructure sized for peak loads that happen twice a year.
This project attacked that problem directly:
→ Multi-cloud GPU placement — AWS for general compute, GCP for TPU access and BigQuery analytics, workload placement driven by price/performance analysis not vendor default → Spot instances for training — 60% cost reduction on interruptible workloads where interruption doesn't matter → Reserved capacity for inference — customer-facing models never compromised for cost → Right-sizing on actual utilization — not theoretical maximums
Result: $1M+ in annual Data Science savings while maintaining cutting-edge ML capabilities.
The freed budget went directly into AI-powered customer features.
That's not cost cutting. That's capital reallocation.
Two pages. Two audiences. One story.
Executive Summary: Where did the money go and did we stay on budget?
- KPI cards: Total spend, total savings, efficiency rate
- 24-month budget variance trend by department
- Department-level performance against targets
- Interactive filters for drill-down analysis
The insight: All 8 departments. Under budget. 24 consecutive months. Remarkable consistency across a volatile optimization program.
CX + AI FinOps Deep Dive: Where is AI spend going and what is it buying?
- Multi-cloud service cost breakdown (AWS vs GCP)
- AI/ML infrastructure costs by service type
- Identified optimization pipeline: $49K remaining opportunity
- The CX connection: Infrastructure savings mapped to customer outcomes
The insight: Every dollar saved in infrastructure has a destination — customer-facing innovation. This page shows the before and after.
AWS — General compute, storage, traditional workloads GCP — BigQuery analytics, specialized ML workloads, TPU access Decision framework — Strategic workload placement based on price/performance analysis, not convenience
- Reserved Instance Management — 40–60% savings on predictable workloads
- Spot Instance Strategy — Training jobs on interruptible capacity, inference protected
- Right-Sizing Program — Continuous monitoring, automated recommendations
- Environment Optimization — Dev/test scaled to actual usage, not always-on
- Waste Elimination — Orphaned resources, unused volumes, stale snapshots systematically removed
All data is fully synthetic — anonymized patterns modeled on realistic AWS + GCP billing structures. No real account info, PII, or proprietary data.
| File | Description |
|---|---|
finops_budget_tracking.csv |
24 months budget vs. actual by department |
finops_cloud_costs.csv |
Detailed costs by provider, service, environment |
To reproduce:
- Download CSV files or connect your own billing exports
- Connect in Tableau Public or Desktop
- Customize filters and annotations for your context
- Publish and share
Analysis period: January 2023 – December 2024 Tools: Tableau Public, Python (data validation)
This framework applies directly to:
- CTOs managing multi-cloud costs at scale without sacrificing product velocity
- CFOs who need infrastructure spend connected to business outcomes not just cloud invoices
- FinOps leaders building the case that optimization enables innovation
- AI/ML platform teams drowning in GPU costs with no framework for right-sizing
- Engineering leaders at product companies where infrastructure efficiency directly funds customer experience
I'm customer obsessed.
That means when I look at a cloud bill I don't see a cost problem. I see capital that isn't reaching the customer yet.
Every orphaned resource is a feature that didn't get built. Every oversized instance is a personalization model that didn't get trained. Every wasted GPU hour is a customer experience that didn't get faster.
I've built data infrastructure from scratch — petabyte-scale lakes, medallion architecture, autonomous AI systems running in live financial markets. I've seen what happens when infrastructure is treated as a cost center versus a strategic enabler.
The difference shows up in the product. Every time.
Customer Problems → Data Infrastructure → Customer Outcomes.
That's not a methodology. That's a philosophy.
Built by Tracy | Apex Engineering LinkedIn | Tableau Dashboard