skip to content
martingale
request a review → describe the build →

production ai systems
for finance

Martingale GmbH is a boutique AI engineering practice for regulated finance.

A few engagements at a time. Chosen for fit.

the gap

Most AI prototypes never survive contact with production. A demo only has to work once. A system in regulated finance has to work every day: reconcile, keep an audit trail, deploy safely, and fail predictably.

Closing that gap is the whole problem, and the only one we take.

the trading day

A live, holiday-aware 24-hour market dial in Frankfurt time. Concentric arcs mark the Tokyo, London, Frankfurt and New York trading sessions; a petrol hand sits at the current Frankfurt time; a market closed for a holiday or weekend is shown dashed and faint and the legend names the reason; a readout states the current Frankfurt time and which markets are open now.

Live holiday-aware global market sessions, Frankfurt time A 24-hour dial with session arcs for Tokyo, London, Frankfurt and New York, a hand at the current Frankfurt time, and closed markets shown dashed. 12 18 00 06

the substrate

Built to one standard. The substrate that lets a finance system be:

  1. 01 instrumented — every decision and signal captured at the seams
  2. 02 evaluated — against a private harness, not opinion
  3. 03 corrected — each caught failure ratchets into the harness and can't recur
  4. 04 improved under control — changes governed, reversible, observable
  5. 05 kept portable — not welded to one model vendor

how we work

Two ways to engage. A fixed-price build, delivered in your stack. An independent review of a system you run, or another shop’s spec.

The build is yours to keep. Monitoring and evaluation after handoff, if you want it.

The review holds a system or a spec to the same standard: whether it will survive production.

Builds are few by design. Reviews are shorter and start sooner.

services

  • independent technical review

    A third-party assessment of whether an AI system or spec will survive production.

  • design & deployment

    The core build, made to run every day under real load, not to work once in a demo.

  • monitoring & evaluation

    Standing evaluation and instrumentation to detect drift or degradation as a live system runs.

  • risk & governance

    Audit trails, contained blast radius, predictable failure: the controls that make a system safe to run under regulation.

the practitioner

A builder and researcher.

Martingale is led by Gregory Wheeler. He co-founded a digital lending marketplace and built its credit-scoring pipeline end to end — independent assessment of loans for institutional investors to fund. He publishes on probability and decision theory, the mathematics that finance runs on.

contact

Describe the work.

What you’re building, or what you’d have reviewed. The constraints it has to hold, and where it stands.