Stop paying for idle compute. Stop burning sprint cycles on infrastructure toil. I find the waste, implement the fixes, and hand you the Terraform — so your team stays focused on shipping product. Free audit. No commitment.
For Series B–D SaaS teams running Kubernetes on AWS.
Your engineers could do this — if cost optimization were their full-time job.
It isn't. They're shipping product, which is exactly where their focus belongs. I step in as the specialist so your team doesn't have to context-switch.
If you're running EKS at Series B+ scale, at least one of these is active in your cluster today.
Industry surveys consistently put EKS clusters well below 50% utilization — while you pay for 100%. Without Karpenter and proper autoscaling, mid-scale deployments carry idle capacity that adds up fast.
Often $10–40K/mo at scale$20K/mo saved = a senior engineer added to your payroll.
Health-check logs, debug output, and high-volume low-signal spans ship to Datadog at full price. Teams typically waste 30–50% of their log budget on data nobody queries.
Typically 30–50% of log spendEvery byte between availability zones is billable. Without topology-aware routing, Kubernetes sends internal traffic across zones by default — and the transfer fees compound quietly until someone looks.
$10–30K/mo at high trafficEvery audit runs the same way. Here's exactly what I do.
Where the money actually goes — pulled from Cost Explorer and attributed by workload, not guessed.
Provisioned vs. real usage from CloudWatch and Datadog, measured at p95 — not point-in-time averages that hide the bursts.
Compute, observability, network — each with a confidence range, so you can decide what to act on without a second opinion.
Ranked by dollars recoverable vs. effort — what to fix first, and what isn't worth touching.
No infrastructure access for the initial scoping conversation — just your spend profile and a quick architecture overview.
A short call to map your architecture and spend profile. No access required at this stage.
You grant read-only, least-privilege access under NDA — no write permissions, no production changes, removable at any time. I run a structured, hands-on review across compute, observability, and network, and deliver a written findings summary either way.
If the numbers justify it, I implement the fixes in coordination with your team — Karpenter, log-pipeline cleanup, topology routing. Terraform and YAML, not PowerPoint. Industry benchmarks put 20–35% savings within reach across the areas I cover; your audit shows what's realistic for your environment.
Zero impact on uptime or performance buffers — we remove the waste, not your safety margins.
Right-sized nodes, Karpenter (spot + bin-packing), VPA request/limit tuning.
Typical: 20–30% of compute spendLog-pipeline filtering, retention tuning, instrumentation cleanup.
Typical: 30–50% of observability spendTopology-aware routing, NAT Gateway optimization, egress reduction.
Typical: 20–50% of transfer costRanges are typical opportunities and vary by setup. Your audit shows your specific numbers.
Tools like Kubecost or Finout show you where the waste is. They don't right-size your node groups or rewrite your log pipeline. I do the work the dashboard surfaces.
Your engineers could do this — if cost optimization were their full-time job. It isn't; they're shipping product. It's the only thing I do.
The waste compounds every month it runs. The audit is free, read-only, and asks almost nothing of your team — there's little reason to wait.
It's not a pitch disguised as a review. You get a written findings summary either way. If I don't find meaningful, recoverable waste, I'll tell you that directly on the call — we part ways, no invoice, no obligation.
Read-only means a scoped IAM role you control and can remove at any time — no write access, nothing deployed to your cluster, and a mutual NDA signed before anything is shared.
I've spent 7 years building and operating production Kubernetes — systems where a misconfigured node group or an unfiltered log pipeline quietly burns tens of thousands a month before anyone notices. The waste is rarely where the team expects: idle daemonsets, health-check cardinality in Datadog, cross-AZ calls nobody ever mapped.
I don't consult from the sidelines. I write the Terraform, tune the Karpenter configs, and clean the Datadog pipelines. You get engineering execution, not a report you have to act on yourself.
K8s Cost Ops does one thing — find and remove that waste, completely.
The three highest-impact areas are compute over-provisioning, observability waste, and cross-AZ data transfer. Right-sizing nodes and enabling Karpenter with Spot instances typically recovers 20–30% of compute spend. Filtering Datadog pipelines to remove health-check noise recovers 30–50% of observability spend. Enabling topology-aware routing eliminates cross-AZ transfer fees that commonly run $10–30K/month at Series B–D scale.
Most Series B–D SaaS teams recover 20–40% of their total Kubernetes-related AWS spend through structured optimization — often $40–120K per year combined across compute (20–30%), observability (30–50%), and network (20–50%). Your free audit shows the exact numbers for your environment before you commit to anything.
Karpenter provisions the exact instance type each workload needs directly from EC2, enables Spot instance fallback, and bin-packs pods more efficiently than Cluster Autoscaler. Teams switching typically see 20–35% compute cost reductions — but right-sizing pod resource requests must come first. Switching without a prior request/limit audit often leads to poor instance selection and higher costs initially.
Almost always log pipeline waste: health-check logs, debug output, and high-cardinality spans that ship to Datadog at full price but generate zero actionable signal. Most teams waste 30–50% of their Datadog spend on data nobody queries. Filtering at the Vector or FluentBit layer before logs reach Datadog typically cuts observability spend by $8–30K per month at mid-scale.
Cross-AZ transfer occurs when Kubernetes routes traffic between pods in different availability zones — AWS charges $0.01/GB each direction. Without topology-aware routing, Kubernetes sends traffic wherever capacity exists regardless of zone, generating $10–30K per month in avoidable fees at high-traffic scale. Enabling topology-aware service routing (available since Kubernetes 1.21) eliminates most of it with a single annotation change.
A 24-hour read-only review using a scoped IAM role — CloudWatch and Cost Explorer access only, no write permissions, removable at any time. The audit covers compute utilization vs. provisioned capacity, observability spend by service, and network cost patterns. You receive a written report with waste by category and monthly dollar values. If no meaningful recoverable waste is found, you're told directly — no invoice, no obligation.
I'll respond within 2 business days to schedule. NDA first. Read-only access, removable anytime. No obligation.
Prefer to talk first? Book a 15-min call →