Skip to content
IRONVEIL

Architecture and security for infrastructure you own.

Ironveil Systems is an Arizona LLC doing architecture, security, monitoring, migrations, and managed hosting — as your ongoing provider or as the team that sets it up and teaches you to run it yourself. Either way you get the configuration, the documentation, and a written account of what each control stops.

Start a conversation See WASP →

01 / WHAT WE DO

Nine areas of work, all of them hands-on.

We work as an MSP directly, and we take escalations from other MSPs — the infrastructure problem that has outlasted two vendors, the migration nobody wants to own, the outage whose cause is still unexplained. Subcontract or white-label, your client relationship stays yours.

Infrastructure architecture

Network segmentation, hypervisor layout, storage and backup design for Proxmox, bare metal, and mixed environments. We write the topology down, including the parts that are compromises, and hand over the diagrams and the runbooks.

Security engineering

Firewall policy, host hardening, patch and release process, key handling, log collection that someone actually reads. Work starts with a written threat model so it is clear which attacks the result addresses and which it does not.

Self-hosted platforms

Standing up mail, identity, version control, monitoring, and web hosting on infrastructure you own. Includes the unglamorous half: restore testing, upgrade paths, and an honest estimate of the operating hours it will cost you.

Managed hosting and migrations

Managed web hosting on our stack or yours, and moves between them: server to server, on-premises to cloud, cloud back to hardware you own. We also move organisations between platforms, such as Google Workspace to Microsoft 365 or the reverse, taking mail, files, calendars and identity across. Migrations run with a rehearsed cutover, a rollback plan, and a maintenance window we commit to in writing.

Systems administration

Day-to-day administration and architecture across Azure, other clouds, Linux and Windows fleets: identity, networking, backup, patching, and cost. We have designed multi-tenant Entra ID estates with delegated business-unit administration — management groups, RBAC, Privileged Identity Management — and documented them so someone else can run them.

GPU and private AI compute

We can place dedicated NVIDIA H200 capacity and isolated AI environments in a SOC 2 Type II and ISO 27001 audited Midwest data center. We architect the environment, set the access and network boundaries, and operate it — your data and models stay inside it.

Governance, risk and compliance

Policy, control mapping, audit preparation, and risk assessment, handled by credentialed GRC practitioners on our team. If compliance work is what you need, say so when you write and we will put you in front of the right person rather than improvising it.

Personal privacy and operational security

For individuals with real exposure: reducing the personal data already public, separating identity from address and phone, hardening devices and accounts, and building habits that hold. Starts with a written threat model, because privacy work costs convenience and you should know what you are buying.

Research and development

Applied work on privacy properties of communication systems, plus the tooling that comes out of client engagements. Results are published when they hold up, and withdrawn when they do not.

MANAGED HOSTING · PRICING

STATIC SITE

$15–30per month

Hosting, TLS, CDN and uptime monitoring.

WORDPRESS

from $40per month

A dedicated hardened VM, filtered outbound traffic, encrypted off-site backups, and updates tested on a copy first.

APPLICATIONS

$15–200per month

Depends on the stack, from a single service to databases and multiple components.

Ask for a quote → Contact us for an accurate quote.

SELECTED WORK

Two recent engagements. Client names are withheld; the figures come from the engagement records.

SOFTWARE HOLDING COMPANY · MULTIPLE BUSINESS UNITS

80+ servers out of a closing data center, one business unit at a time

The data center hosting a shared Hyper-V environment was closing. Its servers ran order entry, license-key services, development environments, web hosting and file storage for several business units, some of them customer-facing with thousands of users.

80+
servers rehomed
18 days
delivery schedule per business unit
2003 → 2022
Windows Server versions, upgraded on arrival

WHAT WE DID

  • Built the Azure hierarchy with a management group per business unit, so each unit’s IT staff administer their own resources and nothing else. Role assignments are made at resource-group level, with just-in-time elevation for tenant-wide roles.
  • Gave each unit its own isolated virtual network, peered only to a management network for bastion access. Network security groups restrict every VM to specific protocols and source addresses.
  • Let each unit choose its destination (its own Azure, AWS or private cloud) or hand the migration to us.
  • Piloted every server with its owner, then repeated the pilot steps against fresh backup data at cutover, followed by acceptance tests, documentation and training.
  • Deployed across availability zones for a 99.99% uptime target, with 14-day instant-restore backups and centrally managed endpoint protection.
Facing something similar? →

NONPROFIT · SCHOOL-BASED COUNSELING · ~110 STAFF

Google Workspace to Microsoft 365, with an identity model a part-time admin can run

The organization had no security groups, no consistent address format, no device policy, and admin rights on everyday mailboxes. Staff handle student and clinical data, mostly on personal phones. Google Workspace had to stay in service for file sharing with partner school systems.

63
mailboxes migrated
164,000+
items moved
0
failed migrations, no data loss

WHAT WE DID

  • Reconciled the staff roster against the live directory before batching. This found mailboxes the roster did not list, which a roster-based migration would have left behind.
  • Ran seven batches, smallest first. The pilot was validated item by item, and the largest mailbox (35,000 items, 15 GB) ran alone.
  • Added every legacy address as an alias and checked each one resolved before moving the MX record, so the domain cutover dropped no mail.
  • Separated admin privilege from mail-receiving accounts, and set up emergency access accounts with credentials stored offline.
  • Twelve role-based security groups, with a single group assigning licenses. Offboarding is one membership change. Unused seats were reclaimed without new purchases.
  • Used app protection instead of device enrollment. Work data stays encrypted inside managed apps, and staff’s personal data is out of the organization’s reach.
Facing something similar? →

CLIENTS

Software companies and operators we have done infrastructure, security, or platform work for.

  • Shoplogix
  • Industrios Software
  • Argos Software
  • Cadre Technologies
  • CDR
  • Comparatio
  • Compusense
  • ASA Automotive Systems
  • DPS
  • Factivity
  • bitec
  • Crescent One
  • Freestyle Solutions
  • A+W Software
  • Friedman Corporation
  • Elm Tree System
  • Datamine
  • ASC Software
  • Xdata Solutions
  • Libra Software Group
  • Logimax
  • DMSpro
  • MegaBus Software
  • Optitex
  • Pace Software
  • NedGraphics
  • infoBATE
  • PMC Software
  • ProShip
  • SpectraQEST
  • SphereWMS
  • Stat-Ease
  • Tecplot
  • Varsity Logistics
  • Viewlocity
  • Kato Integrations
  • Ziiware Inc.
  • Volo
  • Vortx
  • Windflite
  • WinPro X

02 / AI COMPUTE

Dedicated H200 capacity, inside a boundary we build.

We rent NVIDIA H200 nodes in a SOC 2 Type II and ISO 27001 audited Midwest data center. We set the access and network boundaries and operate the environment. Your data and models stay inside it.

What you can rent

Whole dedicated H200 nodes, or a share of one for smaller workloads.

Terms

Hourly, monthly, or on contract.

What we handle

Environment architecture, access control, network boundaries, and day-to-day operation, documented the same way as everything else we hand over.

Ask about capacity Availability and pricing on request.

03 / OPEN SOURCE

WASP

A provisioning system that builds hardened WordPress hosts on Proxmox VE. You give it a host and a site list; it produces segmented VMs or containers, a default-deny firewall policy, signed releases, and backups it verifies by restoring them.

Written for MSPs and small teams who run client sites and would rather not rebuild the same stack by hand every time. It is opinionated about layout and process, which is the point: the same shape every time is what makes an incident tractable at 2am.

MIT licensed. No hosted control plane, no telemetry, no account. Read the code before you run it.

You can also have us run it. Our managed hosting is WASP on our hardware: we handle provisioning, patching, signing keys, and the restore tests, and you get the same segmentation and firewall policy as a self-hosted install. Because the stack is the open one, leaving means taking the configuration with you rather than starting over.

WHAT IT GUARANTEES

  • Each site gets its own network segment; east-west traffic between sites is denied unless a rule names it.
  • Egress is allowlisted per host, so a compromised plugin cannot phone out to an address you have not approved.
  • Releases are signed and version-pinned; provisioning refuses an artifact whose signature does not verify.
  • Backups are restored into a scratch VM on a schedule and the restore is checked, not just the archive.

WHERE IT STOPS

  • It does not audit plugin or theme code. A vulnerable plugin inside an allowed path is still a vulnerability.
  • It does not defend against a compromised Proxmox host or hypervisor-level access.
  • It assumes you keep the signing key off the provisioned hosts. Nothing enforces that for you.
  • It has not been through third-party audit. The threat model is in the repo; read it and disagree with it.

04 / RESEARCH & DEVELOPMENT

Two active programs.

We keep a research line running alongside client work. Both programs below are under investigation rather than for sale, and both are described here at the stage they are actually at.

Neither has been through third-party review. Where we have measurements, they come from our own lab under conditions we chose, which is a weaker claim than it sounds and we would rather say so than imply otherwise.

  1. PROGRAM 01

    HOLLOW

    Security operations tooling that rebuilds its own compromised hosts. Working lab build.

  2. PROGRAM 02

    Privacy-preserving communications

    How much metadata leakage a usable messaging system can shed. Simulation only.

Read the research →

05 / HOW WE WORK

Three rules we hold to.

01

Self-hosted by default

Hosted services are a fine choice when you make it deliberately. We start from infrastructure you own, then add outside dependencies one at a time, each with a reason and an exit.

02

Controls that state their limits

Every control we ship comes with the attack it stops and the attack it does not. A control whose boundary nobody wrote down gets trusted past its design, which is how people get surprised.

03

Nothing you cannot audit

Configuration lives in files you can read and diff. No black-box agents, no vendor console holding the only copy of the truth. If we leave, everything needed to keep running stays with you.

Larger engagements run in phases — envisioning, planning, developing, deploying and stabilizing — adapted from the Microsoft Solutions Framework. Each phase ends with a written artifact: a scope, a design document, a risk assessment, a pilot and acceptance plan, and a handover to whoever operates the system next.

06 / CONTACT

Tell us what you are running.

Describe the environment and the problem. We will reply with what we would look at first and whether we are the right people for it.

Ironveil Systems LLC · Arizona, United States

EMAIL

hello@ironveil.systems

Start an email →

PHONE

(623) 268-7577

SIMPLEX

End-to-end encrypted chat with no phone number, account, or user ID. Use this if you would rather not put the details in email.

WHAT TO INCLUDE

  • What you are running: hosts, hypervisor, clouds, rough size.
  • What is going wrong, or what you want to change.
  • Any deadline or constraint we should know first.