FinOps Engineer Roadmap

Skills you'll master

Hands-on projects

1. 01Enable a raw billing export (AWS CUR, GCP BigQuery billing export or Azure Cost Management export) and query it directly with SQL rather than through the console
2. 02Reconcile last month's billing export against the invoice total and write down every line that did not match and why
3. 03Produce the same monthly figure three ways — on-demand, amortised and blended — and write the one-paragraph explanation of when each is the honest number
4. 04Map one workload's bill down to individual line items, naming the service, usage type and operation behind each charge
5. 05Convert a provider's native billing export into FOCUS columns and document which fields lost meaning in the translation
6. 06Take a service you use daily and write down exactly what triggers a charge — most engineers get storage or egress wrong on the first attempt

Resources

Skills you'll master

Hands-on projects

1. 01Build a showback report that attributes at least 90% of last month's bill to a named team, and list what remains unattributed
2. 02Design a tagging schema with required keys and allowed values, then measure current coverage against it as a single percentage
3. 03Pick one shared cost — a NAT gateway, a logging pipeline, a shared cluster — and implement a defensible split, then present it to the teams being charged
4. 04Enrich raw billing data with an ownership lookup table so every line item resolves to a team without anyone reading a tag by hand
5. 05Write the chargeback proposal you would take to finance, including what happens when a team disputes its number
6. 06Track tag coverage weekly for a month and show whether the trend moved after you told people about it

Resources

Some resources are affiliate links, marked AFF. They cost you nothing extra, we may earn a commission, and a free alternative is listed alongside wherever one exists.

Skills you'll master

Hands-on projects

1. 01Define and publish one unit-cost metric for a real service, and chart it against volume for the last three months
2. 02Find a month where total spend rose but unit cost fell, and write the two-sentence explanation a CFO would accept
3. 03Calculate cost to serve for your most expensive customer segment and compare it against what that segment pays
4. 04Build a pipeline that joins billing data to a business volume metric on a daily grain, so the unit cost updates without manual work
5. 05Pick a service whose unit cost you cannot compute, and write down exactly which missing data blocks it
6. 06Present a unit economics review to someone outside engineering and record which numbers they questioned first

Resources

Skills you'll master

Hands-on projects

1. 01Model a one-year and a three-year commitment for one workload, showing break-even month and the loss if usage falls 30%
2. 02Report current commitment coverage and utilisation as two separate numbers, and explain why optimising only one of them is a trap
3. 03Move a fault-tolerant batch workload onto spot capacity and measure both the saving and the interruption rate you actually saw
4. 04Calculate your effective savings rate across a whole account and compare it against the headline discount the provider advertises
5. 05Write the commitment purchase recommendation you would sign, including the usage forecast it depends on and what invalidates it
6. 06Audit existing commitments for waste — find the ones covering capacity that no longer runs, and quantify the monthly loss

Resources

Skills you'll master

Hands-on projects

1. 01Cut one workload's monthly spend by at least 20% without changing its service level, and publish before-and-after utilisation
2. 02Build an idle resource report — unattached disks, idle load balancers, forgotten environments — and drive it to zero with named owners
3. 03Implement a scheduled shutdown for non-production environments and measure the actual saving, not the theoretical one
4. 04Rightsize a fleet from observed utilisation percentiles rather than averages, and document what headroom you deliberately kept
5. 05Apply a storage lifecycle policy that tiers or expires old objects, then verify nothing that mattered was moved
6. 06Trace a surprising data transfer charge to the architecture decision that caused it, and propose the fix with its cost

Resources

Skills you'll master

Hands-on projects

1. 01Publish a monthly forecast, then score it against actuals at month end and report the variance as a single percentage
2. 02Set budgets with tiered alert thresholds routed to the team that can act, not to a shared inbox nobody reads
3. 03Build anomaly detection that catches a deliberately introduced cost spike within 24 hours, and tune it until false positives are rare
4. 04Rebuild a forecast as a driver-based model — volume times unit cost — and compare its accuracy against straight-line extrapolation
5. 05Investigate a real variance over 15% and write the explanation, separating price changes from usage changes
6. 06Write the escalation policy for a budget breach, naming who decides between paying it and stopping the workload

Resources

Skills you'll master

Hands-on projects

1. 01Run a monthly cost review with an engineering team and record which decisions changed as a result — that count is the only real output
2. 02Get cost data in front of engineers where they already work — pull request, dashboard, chat — rather than in a report they must go and open
3. 03Write the one-page cost brief for a product manager that leads with unit economics rather than total spend
4. 04Identify a workload where the cheapest option was rejected for a good reason, and document the tradeoff so it is not relitigated monthly
5. 05Design a lightweight efficiency target a team can own, and check three months later whether anyone still tracks it
6. 06Facilitate a session between finance and engineering on one disputed number, and write down the vocabulary mismatch you found

Resources

Skills you'll master

Hands-on projects

1. 01Deploy OpenCost on a real cluster and produce a per-namespace cost report that reconciles to the cloud provider's node bill
2. 02Quantify idle cluster capacity as a monetary figure, and decide who pays for it — the split you choose is the interesting part
3. 03Compare cost calculated from resource requests against cost from actual usage, and write down which one you would charge teams on and why
4. 04Find the worst over-requesting workload in a cluster, right-size it, and measure the node capacity reclaimed
5. 05Introduce a node provisioner such as Karpenter or cluster autoscaler tuning, then measure the change in cost per pod-hour
6. 06Build a namespace showback dashboard a platform team can hand to tenants without further explanation

Resources

Skills you'll master

Hands-on projects

1. 01Write a policy that rejects untagged resources at provisioning time, and measure tag coverage before and after it shipped
2. 02Add a cost estimate to pull requests that change infrastructure, so the price appears during review rather than on the invoice
3. 03Automate cleanup of one class of orphaned resource, with a dry-run mode and an audit log of everything it deleted
4. 04Document the failure modes of your own automation, answering what happens if it deletes something still in use
5. 05Enforce a guardrail — instance family, region, or storage class — then track how often engineers requested an exception
6. 06Codify a commitment coverage check that runs weekly and opens a ticket when coverage drifts below target

Resources

Skills you'll master

Hands-on projects

1. 01Run a crawl-walk-run maturity assessment across the FinOps capabilities and publish a prioritised twelve-month plan from the gaps
2. 02Define the five KPIs your practice reports to leadership, with the query behind each one so nobody recalculates them by hand
3. 03Build a self-service cost tool teams use without asking you, then measure usage after you stop promoting it
4. 04Extend allocation beyond one provider to a second cloud or a major SaaS spend, and normalise both into one report
5. 05Write the quarterly executive cost narrative — what changed, what it cost, what you recommend — in one page
6. 06Produce the business case for a FinOps tooling decision, comparing build against buy with the operating cost of each

Resources