Deep AWS Cost Engineering.
We cut your monthly AWS bill with fixes AWS will never suggest.
Every recommendation explains the cause, the fix and the risk
Our engineers work directly with your billing, usage and architecture. Each change is written as a pull request against your infrastructure code, so your team reviews it before anything ships.
A few examples of what we advise on
Savings Plans and rightsizing are worth doing, and most tools will recommend them. Much of the remaining savings sits in the architecture, where AWS and AI tools rarely suggest changes. Every system needs its own answer; these are some directions we often take.
NAT Gateway
Route S3 and ECR traffic through VPC Endpoints, and consider NAT instances where throughput and availability requirements allow. We agree on operations and failover with you before anything changes.
NLB / ALB
Consolidate separate load balancers into a shared ingress, and review cross-zone settings to reduce data transfer between AZs.
CloudFront
Serve egress through CloudFront instead of straight to the internet, and tune cache keys and TTLs to raise the cache hit ratio.
S3
Break down cost per bucket, set lifecycle rules that match how data is accessed, and advise on alternatives when S3 is no longer the economical choice.
Kubernetes (EKS)
Align requests and limits with real usage, improve bin-packing, and consider Graviton or Spot for workloads that suit them.
RDS / Database
Tune queries, indexes and connection pooling before downsizing instances, so savings don’t cost you performance.
Logging & observability
Lower CloudWatch or Datadog spend with sampling and sensible retention, and consider keeping logs in S3 or ClickHouse.
Cross-AZ & data transfer
Find data moving between AZs or out to the internet, then adjust routing, replication and data formats.
This is not a complete list. Every recommendation starts from how your system actually runs.
- 01 · [N] days
Assessment
We analyze your infrastructure and deliver a list of recommendations, each with estimated savings and risk.
- 02 · [N] weeks
Implementation
We ship the changes with your team through pull requests, measuring cost and performance before and after each one.
- 03 · monthly
Continuous
Automation and AI agents watch your costs daily, and our engineers step in when something drifts.
Pricing tied to real results
Alongside fixed packages, we can charge a percentage of verified savings. How savings are measured, and against which baseline, is agreed in the contract.
- Assessment
- [PRICE]
- Implementation
- [PRICE] / project
- Continuous
- [PRICE] / month
- Performance-based
- [X]% of savings
Frequently asked questions
What access do you need?
To start, send us your billing data (a Cost and Usage Report or a Cost Explorer export), or grant a read-only IAM role. We sign an NDA before receiving any data.
Will the optimization affect production?
Every change is written as a pull request against your IaC, with a rollback plan, and ships only after your team reviews it. We measure cost and performance before and after each change.
How is CostKernel different from FinOps or AI tools?
Tools are good at spotting idle resources and recommending Savings Plans. We focus on architecture changes, which require understanding how your system runs and someone accountable for shipping them.
Which clouds do you support?
We currently focus entirely on AWS, along with services around it such as Datadog, Cloudflare and MongoDB Atlas. If you run on GCP or Azure, get in touch and we can talk.
Which companies are a good fit?
SaaS, e-commerce and fintech companies spending [SPEND LEVEL] or more per month on AWS, with an engineering team that runs its own infrastructure.
How is the service priced?
You can choose an Assessment package, per-project fees, a monthly retainer, or a percentage of verified savings. We recommend a model after our first conversation.
Engineering notes
What we learn from working on real AWS bills, written for engineers. Articles are currently published in Vietnamese.
Start with your AWS bill
Leave your details and we will get back to you within one business day to set up a first call. Within [N] days of receiving your billing data, you will have our first recommendations.