What we build

We build AI products and regulated systems.

Most software shops optimise for shipping fast. We optimise for software where that is not enough: AI that acts on real data, money that has to reconcile, controls a regulator will ask about. This is the work we take on.

System Asset tokenization platform · anonymized
Investor onboarding KYC / AML, accreditation checks
Issuance & registry ownership records, 29 jurisdictions
on-chain Token contracts transfer rules enforced
off-chain Owner of record legal registry of truth
fractional ownership · cross-border transfer rules · audited records
§ I. What we take on

Six kinds of work, one standard.

Some of these run for weeks, some for months. All of them get the same treatment: a written scope, a fixed price, and the failure cases handled first.

If your project doesn't match one of these six, we'll say so on the call rather than take the work and figure it out as we go.

  1. 01
    AI products and agents
    Agents, retrieval and evaluation built around your data, with spending limits and permission checks the model can't talk its way past.
  2. 02
    Tokenization and real-world assets
    Platforms for fractional ownership and on-chain records, with the cross-border compliance that keeps them legal. We have taken one to $216M of tokenized real estate across 29 countries.
  3. 03
    Trading and settlement
    Clearing and settlement infrastructure that holds under both load and audit, wired into the rails and brokerages you already use.
  4. 04
    Payments and money movement
    Idempotent, reconciled money flows with the AML and KYC controls a regulator expects to see, built in from the first commit.
  5. 05
    Identity and onboarding
    KYC and document handling under GDPR and PDPL, with government-ID integration and data residency where the law demands it.
  6. 06
    Rescue and rebuild
    When a rushed or AI-generated codebase can't carry the next stage, we rebuild the core the way it should have been, with the negative tests the generator never wrote.
§ II. How engagements work

Fixed price, scoped on a call, no hourly billing.

Every engagement starts with a scoping call. We write the deliverable down, agree a fixed price, and build it. There's no hourly meter running, and no open-ended "discovery phase" to pad the bill.

We work in your repo, under your review. You get pull requests and the tests that match what we changed, including the negative tests that prove the controls hold. Nothing lands until you've merged it yourself.

Engagement length depends on the shape. Rebuilds and hardening run one to four weeks. Greenfield builds run two to eight. Anything longer gets scoped in phases, each one a place you can stop.

§ III. When this isn't right

What we say no to.

We turn down work outside these six. Consumer social, games, research for its own sake: not our domain, and we'll say so on the call instead of learning on your budget. If AI is the wrong answer to your problem, we'll say that too.

We also turn down briefs too vague to scope. "Make it better" is not a brief. If the call surfaces that the project isn't ready to build yet, the honest answer is sometimes to start with an audit of what you have, sometimes to point you to a different partner.

Have something in mind? Send a paragraph or two about what you need. We'll reply within a day with whether it's a fit, and what a scoping call would cover if it is.

Tell us what you're building