Your dashboards found the waste, but nobody is implementing it.
FinOps implementation for AWS teams. A senior engineer whose only job is closing the cost tickets your engineering team keeps pushing down the backlog. It's not another dashboard, and it's not another report.
Savings ledger
One FinTech client. Nine changes. About $1M a year in savings.
- Databases for a cancelled client
- $378,000
- RDS rightsizing and RI coverage
- $161,000
- EBS on the wrong tier (code bug)
- $156,000
- Aurora Serverless on a flat load
- $125,000
- ElastiCache rightsizing, Redis to Valkey
- $86,000
- Four smaller changes
- $82,000
Total
$988,000/yr
Hello, this is Cristian Magherusan-Stanciu. I've spent 12 years in AWS cost optimization in different capacities, some of them at AWS itself, in the EC2 team, as a Specialist Solutions Architect for Spot and Graviton. I built AutoSpotting, the open-source alternative to commercial AWS cost optimization tooling like Spot.io, used by teams at Samsung, Expedia and Mozilla.
I help teams spending upwards of $100k/month on cloud. I'm specialized on AWS, and cover Azure and GCP through my specialist network.
The FinOps Foundation has a working group dedicated to one problem: getting engineering to take action.
Have a look at your ticket queue and count how many say unused RIs, oversized instance, orphaned volume. Filed months ago, still open, still costing you money every day they sit there.
You're not imagining this. Around 40% of FinOps practitioners say the same thing: they can't get engineering to act. The hard part isn't finding the waste, it's getting it fixed.
The group is called "Encouraging Engineers to Take Action." A whole working group, to get engineers to do something they already know needs doing.
Top challenge cited by FinOps practitioners: "getting engineers to take action"
Source: FinOps Foundation, State of FinOps.
- Jan 15Unused RDS, est. $31.5k/mo
- Feb 15Unused RDS, est. $31.5k/mo
- Mar 15Unused RDS, est. $31.5k/mo
- Apr 15Unused RDS, est. $31.5k/mo
- May 15Unused RDS, est. $31.5k/mo
- Jun 15Unused RDS, est. $31.5k/mo
- Jul 15Unused RDS, est. $31.5k/mo
Seven months, same alert, still open. That's $220,500 out the door while everyone agreed it was a problem. This one's real, by the way: it's the top line of the ledger further down.
Here's the part that doesn't make it into the survey. It's not laziness, and it's not incompetence. Your engineers are doing exactly what they're paid to do. They're measured on features shipped, not dollars saved, so every cost ticket you file walks into planning and loses to the roadmap. Every time, in every company, because that's how the incentives were built. Nobody's promotion packet says "decommissioned the forgotten database", so it doesn't get decommissioned.
You already know all this. You built the dashboard, wrote the report, filed the ticket, followed up, escalated, filed it again. The waste is still running, and somehow you're still the one explaining to the CFO why the number hasn't moved.
You've got all of the responsibility and none of the capacity to act on it.
You have heard all of this before
And you were right to stop believing it.
"Our dashboard gives you better visibility."
You've got visibility. That's not what's broken.
"We will produce a cost optimization report."
You can write your own reports. More reporting isn't what's missing.
"AI-powered cost optimization."
Nobody at your scale is letting an unsupervised model loose in production, and they're right not to.
"Our consultants will review your infrastructure."
You've had consultants. They produced a PDF and left, and getting engineering to implement it was still your job.
"We saved company X $500k."
Every vendor claims this. Nobody ever says who actually implemented it.
"Bring in someone with a bulldozer mentality."
You're still fighting structural incentives with force. Engineers are measured on features, and no amount of pushing changes what they're rewarded for.
Every one of them tries to get engineering to act, whether through better data, better reports, better persuasion or brute force. They all stop at information, and information was never the problem. What's missing is somebody to do the work.
You don't need a better business case. You need implementation capacity.
FinOps implementation is a senior engineer dedicated to the cost optimization work your engineering team won't prioritize: the config changes, the rightsizing, the migrations. So the savings your dashboards have been pointing at for months actually show up on next month's bill. I'm that kind of engineer.
You get your own engineer
Not borrowed from the engineering team and pulled onto a feature next sprint. Someone whose priority is your priority, so you stop depending on engineering's goodwill to get your job done.
The waste you already found gets cut now
Not next quarter when engineering has bandwidth. The person implementing doesn't compete with the roadmap, so there's nothing to lose to.
Your existing practice gets more valuable
This doesn't replace your FinOps practice. Your dashboards and reports stop being the end of the line and become the front of a queue somebody actually works through.
Plus what your tools never surfaced
I bring my own battle-tested tooling, built on 12 years of pattern recognition across many clients, that finds the advanced optimizations most commercial dashboards miss.
The savings you're measured on finally show up, and the job stops being the one where all you get to do is point at problems.
How it works
No transformation program, no lengthy assessment. Just a scoped project with a specific set of implementations and the savings named up front, then the work.
- 1
A call, with the paperwork already signed
NDA up front. A short call to walk through your stack, your existing analysis, and which tickets have been stuck the longest.
- 2
Read-only access
Scoped, read-only access to the AWS account. No standing admin, minimal IAM, and I only ask for more per change, when the work actually needs it.
- 3
We start from your analysis, then go deeper
Your dashboards and tickets are the starting point, not something to redo. We work through what you've already identified, then run our own tooling over the account to find what it didn't surface.
- 4
The changes get made
I make the changes myself: config changes, instance migrations, resource cleanup, and small pull requests that are trivial to review. I pick changes that need zero or close to zero involvement from your engineers, because waiting on engineering is the bottleneck this whole approach exists to get around.
- 5
It shows up on the bill
Measured per change against your billing data from before it landed, over an attribution window we agree up front. The effect usually shows up on the same month's bill. Each week you get a short note of what landed and what it saved. It's a record of what got done, not a list of things for somebody else to do.
We work with your security and procurement process. Everyone who touches your account is under NDA, and we're happy to sign your MSA and DPA, go through your security reviews, and provide references under NDA. We can usually bill through AWS Marketplace, which puts this in the cloud budget you already have and tends to skip a separate procurement cycle for the first project. Whether that route is open depends on the listing, so we confirm it with you rather than assume it.
$988,000 a year. Nine changes. One engineer.
One FinTech client, ~$3M/year AWS spend under an EDP. Every line is a change that actually got made, not a recommendation that got filed.
Savings ledger
| Databases still running for a cancelled client | $378,000 |
| RDS rightsizing, then Reserved Instances to cover it | $161,000 |
| EBS volumes on the wrong tier, from an upstream code bug | $156,000 |
| Aurora Serverless left running on a flat workload | $125,000 |
| ElastiCache rightsizing, then Redis to Valkey | $86,000 |
| Unattached EBS volumes and old snapshots | $27,000 |
| RDS audit log configuration | $23,000 |
| Cross-region RI conversion | $18,000 |
| CloudWatch Logs to S3 | $14,000 |
| Total | $988,000/yr |
None of it needed a slot on engineering's calendar. The engagement is still going and has since passed $1.1M/year in delivered cost avoidance.
E-commerce SaaS, US
~30%
off the bill, ~$168k/yr
A bill growing with the business, from $30k to $50k/month. Cut to $36k/month in 3-4 months across RDS, compute, ElastiCache, S3 and EBS.
Every engagement is under NDA, so client names and identifying details are anonymized; references are available under NDA. More detail on the case studies page.
Why no FinOps team already has this
The reason is structural, and it's worth saying plainly: most senior engineers don't want this work. Unglamorous config changes and migrations don't ship features, don't go on a resume, and mean reporting to the people the engineering org privately calls bean counters. So the role doesn't exist, and the tickets stay open.
I ended up making it my craft instead. My actual job is building cost optimization tools, so I'm a builder like any other engineer, and the implementation work is how what I build gets used. That's why I can stand doing this when nobody else will. It isn't grunt work to me, it's how my tooling gets deployed.
I left AWS because I got bored of PowerPoint.
The other reason the tickets never close
Most of what's left after the headline items is small. A few hundred dollars a month each, low risk, no migration project. On their own, none of them justify pulling an engineer off the roadmap for half a day to research, execute and verify.
That's exactly why they never get done, the arithmetic just doesn't work. It works for us because the tooling is already built, so what costs your engineer half a day costs us half an hour.
Forty of those at $250 a month each is $120,000 a year that no dashboard is ever going to get implemented for you. One piece of that tooling, my EBS Optimizer, is available on its own. The full scope of what gets changed is on the AWS cost optimization page, and the shape of the wider engagement under FinOps consulting.
On the tooling: what teams say about AutoSpotting
“I think AutoSpotting has by far the best approach to utilizing spot instances that I've seen. With AutoSpotting we're pretty much able to run our whole stateless production workload on spot instances.”
Andreas Sundström
DevOps & Development Director, OhMy
“Thanks to AutoSpotting we're saving more than 50% of our server costs on AWS.”
Jan Čurn
CEO, Apify
“Thank you for such an amazing tool, you have the best open source project that I've ever had the pleasure to use.”
Kevin
DevOps Engineer, 3M
No AI touches your production account. Ever.
I use AI for analysis, it lets one engineer cover a lot more ground. But every production change is manual, verified and specific.
No AI agent gets access to your account, and nothing reaches production that I haven't written and checked myself. The amounts at stake are too large and the changes too context-dependent for anything else, and you wouldn't sign off on it either.
How we're engaged
Start with one scoped project. If it works, it continues.
Scoped project
The beachhead: small in hours, large in savings
- A specific set of implementations agreed up front
- Expected savings named per change before we start
- Flat fee, no open-ended engagement
- Proves the model without a procurement cycle
Monthly retainer
Ongoing implementation capacity for your FinOps practice
- An agreed block of engineering hours each month
- Priced case by case, against your scope and what you need
- Your queue, worked through in your priority order
- AWS Marketplace billing, no new budget line
Share of savings
For teams who would rather pay only out of results
- A share of what each change measurably saves
- Billed per change for its first 12 months, then it retires
- Measured automatically from before-and-after billing data
- If a change saves nothing, it isn't charged for
Most teams start with a scoped project, because it's the fastest way to find out whether this works without a budget conversation. All three can usually be billed through AWS Marketplace, which puts the money in the cloud budget you already have rather than needing one of its own.
If you would rather pay out of the savings
Some teams prefer their spend to track results rather than hours. We also work on a share of the savings, typically 10% to 30%, sized to the footprint under our scope, with larger footprints paying a lower share.
Here is how a 20% share played out on a real engagement, change by change.
The same FinTech engagement as the ledger above: ~$3M/year on AWS, where the changes that landed now save about $92,000 a month.
~10%
of the savings we deliver while billing
~$1.1M/yr
you keep, for good, once cuts retire
Each change is billed at the agreed share for its first 12 months, then retires: from then on you keep 100% of that change's savings. It is invoiced monthly and measured automatically from your before-and-after billing data, per change. Across the whole billing period it works out to about 10% of the savings delivered, then it goes to zero while the savings stay. We can bill through AWS Marketplace and will agree a cap if you want one.
About LeanerCloud
I founded LeanerCloud in 2022, after three years at AWS where I worked in the EC2 team as a Specialist Solutions Architect for Spot and Graviton. I've spent 12 years in AWS cost optimization in different capacities, and I created AutoSpotting, the main open-source alternative to commercial AWS cost optimization tooling like Spot.io.
I spent years on the other side of this problem, waiting on engineering teams to implement work that never reached the top of a sprint. The whole approach here is built around not needing them. I run every engagement myself, and bring in a handpicked network of freelance specialists with decades of combined AWS experience when the work calls for it.
I'm specialized on AWS, and cover Azure and GCP through specialists in my network. We're always open to talented cloud professionals interested in joining us.
We often share insights and learnings through our podcast and YouTube channel.
From the blog
Real AWS cost optimization work, with the numbers. Anonymized by scale and spend, every figure measured against the bill before the change.
RDS
Finding and cutting unused and idle RDS databases
Zero-connection databases across dozens of instances and five regions, and over $50k/month of savings potential in a single account.
Aurora
Converting Aurora Serverless to provisioned for cost and performance
A conversion that cut database cost from $450 to $100 a day and improved query latency at the same time.
ElastiCache
ElastiCache cost optimization: rightsizing, Valkey, and reserved instances
Cutting cost over 6x across 200+ clusters, and why rightsizing before committing matters.
Serverless
Moving bursty Fargate and EC2 workloads to Lambda
For spiky web and API workloads, moving to Lambda behind CloudFront often cuts compute cost 10x or more.
S3
Slashing S3 costs with the right storage class and lifecycle rules
Old logs in S3 Standard quietly cost thousands a month. A few lifecycle rules cut one bill by over 90%.
Networking
The cross-AZ data transfer costs hiding in your database traffic
Instances talking to databases across an AZ boundary cost $20k/year here. Finding it by tag, fixing it with a same-AZ replica.
CloudWatch
Cutting CloudWatch Logs costs hiding in a few log groups
A single VPC flow log group cost $8k/year. Retention and log class changes on a handful of groups fixed it.
Send us the ticket that has been open the longest
No pilot program and no new dashboard. Send us the one that has been open longest and we'll come back with what it takes to close it, as a scoped project with the savings named up front.
Frequently Asked Questions
Common questions about working with LeanerCloud
We already have dashboards that show us where the waste is.
Can't our own engineers handle the implementation?
We tried consultants before and they just produced reports.
What about AI-powered optimization?
How do you implement without disrupting engineering?
What if we're already doing some optimization internally?
What does it cost?
Do we need to find a new budget for this?
What size client do you work with?
How much access do you need?
How do you handle security and procurement?
How are the savings measured?
What if a change causes an incident?
Which cloud providers do you cover?
What if we already bought commitments?
Why not just buy a Savings Plan?
Is this just about cutting costs?
What if the bill doesn't drop because we're growing?
What happens when you leave?
How do we get started?
Still have questions?
Send us the ticket that has been open the longest, and we'll tell you what we'd do about it.