Systems we've built and run
The honest way to judge a young firm is by the systems its people have already shipped and run: these six ventures are the founders' own, built and operated end to end.
Building a company teaches what an engagement rarely does, because none of the consequences transfer to anyone else: the security review, the regulator, the outage and the customer who leaves are all yours to answer for. The six ventures below were built under those conditions, across healthcare, security, payments, legal work, social software and network engineering. Some are platforms operating under regulation, one is a working prototype, and one is an engineering firm operating today. Read each for the bar it had to clear, since that bar is what a client is buying when the same people build for them.
-
Miravio
Healthcare SaaS
A clinical operations platform for behavioural-therapy practices in the United States.
Miravio holds the operations of a behavioural-therapy practice in one platform: clinical data, scheduling, billing and the communication that keeps a family informed, rather than separate systems a practice has to reconcile. All of it is handled as regulated clinical data, so access and disclosure are architectural decisions taken at the start rather than settings adjusted later. Each practice is isolated from every other at the data layer, which is the bar a platform must clear before a clinician will put patient information into it, and interoperability with the wider healthcare estate was engineered in rather than bolted on. It is the clearest example of how the partners build under regulation: constraints first, features second.
What this evidences
Regulated clinical data, per-practice isolation and healthcare interoperability are platform problems before they are product problems, and Miravio is where we solved them.
-
Antenna
Open-source SaaS
A social media studio for organizations that run many brands across many networks.
Antenna is the studio an organization runs when one team is accountable for many brands across many networks. Publishing and analytics belong to a single system, so what goes out in each brand's name and what came back from it are read in one place rather than pieced together from several. The software is open source and self-hostable, so an operator can inspect what it does, run it on infrastructure they control and keep custody of the accounts and data inside it. Its AI agents work the same tools the people do, rather than sitting in a separate automation layer bolted to the side. The standard it sets is ownership rather than dependence.
What this evidences
Antenna is a system that speaks to an organization's audience in its own name, with AI agents working the same tools the people do.
-
Redwipe
Cybersecurity
An attack-surface management platform for security teams.
Redwipe watches the part of an organization that faces the internet and tells a security team what is actually exposed. Discovery runs continuously rather than as a point-in-time exercise, because an external estate changes without warning. Findings are verified, and the verification proves a risk is real without harming a live system, which is the line separating an assessment a security team can authorize from one it cannot. Reporting maps to SOC 2, ISO 27001, HIPAA and PCI-DSS, so the output answers the question an auditor asks rather than only the one an engineer would. That is the instinct the partners carry into client work: prove the risk, respect the estate, report in the language the organization is judged in.
What this evidences
Redwipe is the security discipline stated as a product: verify without harm, and report against the standards an organization is actually audited on.
-
Pays
Agentic fintech
A financial operating system for AI agents, built as a working prototype.
Pays is a working prototype of the financial plumbing software agents will need once they begin transacting on an organization's behalf. Every payment is bound to an auditable spend policy, and that binding sits in the first line of the architecture rather than in a control added after money is already moving. The consequence is that the record of what was authorized, by which agent and within which limits, exists by construction, because no path through the system avoids a policy. It is presented as a prototype because that is what it is. What it demonstrates is a habit rather than a product: where money or obligation is involved, the audit trail is designed before the feature.
What this evidences
Pays shows what we mean by auditable finance work, where a spend policy and its record are part of the architecture rather than a later control.
-
Perdais
LegalTech
A UK Global Talent visa platform for candidates, attorneys and case teams.
Perdais carries a UK Global Talent visa application through the three parties it touches: the candidate assembling evidence, the attorney advising on it, and the case team running the process. Each sees exactly what their role allows and nothing beyond it, which matters because the file holds the documents that decide a person's case. Custody of those documents is tracked and an audit trail records what happened to each of them, so a disagreement about the record can be settled from the record. Role-based custody is easy to describe and unforgiving to build, since every feature added afterwards has to respect it. Perdais is where the partners held that discipline in a setting where getting it wrong is not recoverable.
What this evidences
Perdais is software built around who may see what, with document custody and audit trails that have to hold across three different kinds of user.
-
Zilk
AI & network engineering
An AI, cloud, network and security engineering firm, operating today with 450 defined services.
Zilk is the venture operating today, an engineering firm whose people run AI, cloud, network and security work for other organizations. Five engineered products sit alongside the services: network monitoring, multi-vendor network operations, security assessment reporting, charge automation and multicast troubleshooting. Multi-vendor is the demanding word there, because an operations tool that handles only one manufacturer's equipment does not survive contact with a real network. An agentic AI operating layer for network and security teams has been announced, built on the AI already running inside those products rather than on something written for the announcement. Zilk is the proof that the partners can operate a system continuously, not only build one and hand it over.
What this evidences
Zilk is a live operation rather than a delivered project, which is the difference between building a system and keeping one right.
What this is not
These six are the founders' own ventures, not client engagements, and we will not describe them as anything else. Caltor is young and does not publish client case studies, so the evidence we can honestly put in front of you is what the partners built at their own risk and under their own names. That is a narrower claim than a client reference, and a considerably harder one to stage.
Talk to a partner
Each venture names the capability it evidences, and those pages set out what the same engineering looks like inside a client's business.