Eight H100s, thirty percent busy, billed at the full rate.

1. MLOps

   GPU hours nobody is watching

   Rarely measured

   “Your job is running. Your GPUs are idle.”

   Teams watch whether a job is alive. Nobody watches accelerator utilisation. So eight H100s sit at 30% for a fortnight and the bill looks completely normal.

   See the arithmetic
   • Runs under 35% utilisation get flagged as starved, not compute-bound
   • Only the idle share is priced as recoverable
   • Spot runs without checkpointing, where one preemption bins the epochs you paid for
   • Endpoints under 25% utilisation, holding a GPU warm between requests
2. AgentOps

   LLM tokens billed for answers nobody read

   Rarely measured

   “A blocked run costs nothing. That is the whole saving.”

   Cached tokens cost a tenth of fresh ones. Most teams have caching off, or a prompt prefix that shifts just enough to miss it every time.

   See the arithmetic
   • Every run metered against the catalog. Fresh, cached and output priced apart
   • Frontier models doing work a cheaper tier handles at 96% success
   • Caching off, or hitting under a quarter of the time
   • Looping and timed-out runs that billed in full for nothing
   • Guardrails run before billing. A blocked run is a clean saving
3. Cloud accounts · Kubernetes

   Idle cloud and oversized clusters

   “The familiar one. Real money, but everybody already sells it.”

   Unattached volumes. Oversized instances. Node pools still scaled for last year's launch. Every finding arrives with the change attached.

   • Read-only connect to AWS, Azure, GCP, DigitalOcean and OCI
   • Cost and utilisation per resource, on the schedule you set
   • Findings that carry the fix, not just the observation
   • Scale, cordon and drain node pools from the console
4. AIOps

   Alert noise that costs engineers, not dollars

   “One outage should page once. Not three times.”

   Every metric is measured against its own rolling baseline, so a signal has to be unusual to register. Related anomalies collapse into one incident with a probable cause attached.

   • A rolling baseline, not a threshold nobody tuned
   • Anomalies inside five minutes become one incident
   • Playbooks propose the fix, or apply it with the same audit trail
   • Automation can only do what an operator could

Arithmetic on our own accelerator price list, using the formula wasteReport() runs in the product. It is not a measured customer outcome. Rupees at ₹83 to the dollar.

$ coverops waste --domain all   scanning 2 accounts - 2 clusters - 5 models - 4 agents   MLOps     churn-propensity  8x A100-80GB  31% util   $8,412/mo  AgentOps  incident-summarizer  prompt cache off        $327/mo  Cloud     aws-prod  14 unattached volumes  34d idle    $211/mo  AIOps     3 remediations pending approval                   -   total recoverable   $8,950/mo  coverops apply --finding <id>   opens the pull request

One command across all four domains. Every finding carries the money and the fix, and exits non-zero when a budget is breached, so it drops straight into CI.

Your account, your resources

Everything lives in your AWS or Azure account: your billing, your regions, your policies.

A scoped role, no stored keys

Access is a role you create, with a boundary you set. We hold no long-lived keys to lose.

Your audit log sees everything

We act through your provider's API, so CloudTrail and Azure Activity Log record what we did, when, and as whom.

Evidence an auditor will accept

Change history, approvals and policy results export as audit artefacts. CoverOps is not a certification and does not pretend to be.

1. Coming in

   We import what you already run. Nothing is rebuilt, and you see the diff before anything changes.

2. While you use it

   Every change lands as a pull request in your repo. No hidden console state, no proprietary format, no resource only we can describe.

3. Leaving

   Revoke the role and cancel. Your repo still describes the estate, your pipelines still run, and traffic never notices.

Our test: if CoverOps disappeared overnight, your production environment should not notice. If that ever stops being true, we built the wrong thing.

Strongest on that line

Half to start, half on delivery, both on a GST invoice. Read-only unless you tag a resource for us to change. Your cloud bill stays in your own account, and we never mark it up.

How an audit runs

1. BookPick an audit. We confirm the scope in writing and send a GST invoice for half the fee.
2. ConnectYou apply the read-only role CoverOps generates. Removing it ends our access at once.
3. AnalyseCoverOps reads the estate. A person checks every finding before it goes in the report.
4. Hand overThe readout, the report as a PDF in ₹ and $, then the invoice for the balance.

What each audit covers