Operator, not advisor
I ship with your team. I write the policy, wire the GRC platform, open the pull request for the IAM fix, and sit in the incident channel. Recommendations you have to implement yourself aren't deliverables.
rmrfs exists because too many companies get stuck between two bad options: a full-time CISO they can't afford yet, and a consulting firm that hands over a PDF nobody implements. This is the third option. A senior operator who joins your team part-time, ships the program with you, and leaves it running.
I've spent 15+ years building and securing production systems across cloud infrastructure, site reliability, security, and governance, risk and compliance. Today I run all four of those functions as an engineering manager at a B2B SaaS company that sells to enterprises and government.
From the inside, that has meant leading SOC 2, ISO 27001, FedRAMP and IL4 programs from scoping through audit. Answering the questionnaires that gate large deals. Explaining risk to a board in language that gets a decision. And writing the Terraform at 2am when the incident is live. None of this is theory from a checklist.
I also build internal AI agents and tooling for business teams, so I understand AI risk from the builder's side: what an agent can reach, where prompts get injected, and what happens when nobody owns the model inventory.
rmrfs is what I'd have wanted at every company I've helped through an audit: one accountable senior person, part-time, who has done it before.
They will cost me some engagements. Fine. They are also why the ones I take on finish with a program that works.
I ship with your team. I write the policy, wire the GRC platform, open the pull request for the IAM fix, and sit in the incident channel. Recommendations you have to implement yourself aren't deliverables.
A program that only works while I'm on retainer is a failure. Every control gets a named owner on your side, every process gets a runbook, and the handover is planned from week one.
Type II needs an observation window. FedRAMP takes longer than anyone wants. I'd rather lose the deal in the first ten minutes than in month three because I told you what you wanted to hear.
Four concurrent retainers, maximum. That's the number where I still know your environment by heart and reply within a business day. When I'm at capacity, I say so instead of subcontracting you out.
The same four moves whether the goal is a SOC 2 report, a FedRAMP boundary, or an AI governance program. The order matters more than the framework.
Read-only access to cloud, identity and source. Interviews with the people who run things. Contracts and customer commitments on the table before any control is discussed.
A 90-day roadmap with owners, dates and cost. Cheap, high-impact fixes go out immediately. The rest is sequenced around the audit date or the deal, not around a template.
Policies in your voice. Controls wired to real systems. GRC tooling that collects evidence on its own. Working sessions with your engineers, not documents thrown over the wall.
Audits passed, questionnaires answered, vendors reviewed on a cadence. Then the handover: runbooks, owners, and a monthly rhythm that survives without me on the call.
No "dozens of audits" claims. These are the frameworks I've led or run programs for as the vendor, not the auditor. If yours isn't here, I'll tell you and point you to someone.
rm -rf deletes recursively and doesn't ask twice. Every engineer has a story about it. Used carelessly, it takes down production. Used with care, by someone who knows exactly what's in the directory, it removes what shouldn't be there and leaves the rest untouched.
That's the job. Find the stale admin access, the unsigned policies, the unanswered questionnaire, the risk nobody owns, and delete them deliberately. Then hand you a clean tree.
The tiers follow the same grammar. rm for Advisory, rm -r for Managed, rm -rf for Embedded. More flags, more ownership.
Thirty minutes with me, not a sales team. If rmrfs isn't the right fit I'll say so and point you to someone who is. If it is, we'll leave the call with a scope and a start date.