AMA AromaCleaning
Person working at a laptop — Aroma Systems hero, not a data-center rack

Door 2 — Software

Aroma Systems

Cloud junk removal, AI/ML hygiene, and engineering cleanup. Scan first. Approve before anything is deleted.

Book a DemoNeed a space cleaned?

Safety rule

Scan first. Approve next.

Cloud deletion is dangerous. Unknown resources stay tagged do not touch. We start where the bill is — AWS, GCP, Azure.

Checklist used to inventory what exists before any cleanup — look, don’t delete yet
Read-only inventory before any cleanup.
Dashboard review standing in for an approved delete list
You sign the delete list. Then we act.
Signed after-report: what was removed, kept, and the cost delta
After-report: removed, kept, cost delta.

Menu

What we actually do

Same method as a physical deep clean: inspect, isolate junk, clean, verify. Physical crews are on Aroma Clean.

Colleagues reviewing an account list before any cloud cleanup

Always first — read-only

Cloud Dust Scan

A junk map of the connected account. We show what looks idle. We delete nothing.

Included

  • — Inventory across connected AWS, GCP, or Azure accounts
  • — Idle compute and GPU, unattached disks and snapshots
  • — Leftover environments and obvious orphans
  • — Rough monthly waste estimate
  • — Risk tags: safe to review / do not touch / unknown

Not included

  • — Deleting or stopping anything

Best for: Any team that wants a savings range before they approve a cleanup.

Deliverable: Junk map + savings range + recommended next package.

Laptop dashboard used while reviewing an approved cloud reset list

Approved removal

Deep Cloud Reset

The analog of a Deep Reset for infrastructure. You approve the list. Then we clean.

Included

  • — Everything in the Dust Scan
  • — Removals from the approved list
  • — Tagging leftovers
  • — Basic budgets and alerts
  • — After-report: removed vs kept, cost delta, risk notes

Not included

  • — Full architecture rewrite
  • — 24/7 SRE coverage
  • — Production migration without a separate project
  • — Silent destroy — there is no silent destroy

Best for: Accounts that already have a scan and a signed delete list.

Deliverable: Written after-report. Client approves the delete list first.

Written after-report for a recurring Stay Clean Cloud review

Recurring

Stay Clean Cloud

Junk comes back if nobody watches. This is the monthly (or weekly) hygiene pass.

Included

  • — Scheduled junk watch
  • — Drift of idle resources
  • — Short report each cycle
  • — Optional small cleanups inside an agreed cap

Not included

  • — Open-ended rebuilds inside the retainer

Best for: Teams that already did a Reset and do not want the clutter back.

Deliverable: Recurring report, same proof standard as a Reset.

Unused stock in storage — leftover models and jobs that should be retired

AI / ML hygiene

Model Closet

We use AI to find mess. We use engineering to remove it. This offer retires leftover ML.

Included

  • — Inventory of models, endpoints, training jobs, datasets, experiment stores
  • — Flags for unused, duplicate, or stale assets
  • — Retire plan with approval
  • — Cost of idle inference called out

Not included

  • — Training a new production model from scratch (separate engineering project)

Best for: Teams that tried AI and left the lights on.

Deliverable: Closet inventory + approved retire list.

Tablet booking flow for AMA Flow jobs and client updates

Software for cleaning operators

AMA Flow

Booking, job checklists, photo proof, and simple client updates — the ops stack we use on Door 1.

Included

  • — Booking plus job checklist plus photo proof
  • — Simple client updates
  • — Optional intake-to-scope assist (rooms and photos → suggested package)

Not included

  • — A generic CRM rewrite with no cleaning workflow

Best for: Crews that want proof of work without becoming a software company.

Deliverable: Setup plus monthly, or bundled in Operator Studio.

Engineering workspace for a scoped pipeline or integration cleanup

Scoped

Engineering project

Integrations, a specific pipeline cleanup, internal tools, or “what is this bill?” automation.

Included

  • — Written scope before work starts
  • — Cleanup of a named system — not an open retainer by default

Not included

  • — Unbounded “do some AI” with no named deliverable

Best for: When the junk is in one pipeline, one integration, or one internal tool.

Deliverable: Scope document, then the build.

Laptop ready for a Cloud Dust Scan request

Start with a scan

Read-only. No deletions. We map junk, tag risk, and tell you whether a Reset, Stay Clean, or Model Closet is next.

Book a Demo