A project to build with System One models

Suspicious transaction triage, fine-tuned on your own cases

Give every payment a calibrated probability of being suspicious, fine-tuned on your past investigations, and route the risky ones to analysts.

Security and fraudAdvanced4 stepsSuggested by @biplov

The decision

A noul question, "this transaction needs review", over a state built from the transaction and a short account history. It is written as text or JSON: amounts, countries, merchant category, device changes and recent velocity.

How to build it

  1. Export past transactions with their investigation outcomes. Those are your labels.
  2. Fine-tune Laya on them in System One Studio (systemone run studio). It scores the fine-tune against the base on held-out rows before you ship.
  3. Serve it next to your payment flow, and send everything above your threshold to the analyst queue.
  4. Watch precision at the threshold weekly. Fraud patterns drift.

Make it better

Keep a rules engine for the hard constraints, such as sanctions and amount limits, and use the model for the grey zone the rules can't describe.

More projects to build

All projects

Security and fraudIntermediate

A phishing check that warns before anyone clicks

Ask whether each email you open is a phishing attempt, in a browser extension or a mail add-in, with a small model running on the same machine, and show a warning with the probability before anyone clicks a link.

5 stepswith julia-1, laya

Tools and appsAdvanced

A browser agent that fills in web forms

Decide, for each field on a form, whether to type a value, tick a box, click or skip, using a model small enough to run in the loop.

4 stepswith cua-s1-forms

Robotics and edgeAdvanced

A robot's next action, decided on the device

Choose a robot's next action (pick, skip, move, stop) from a text description of what its sensors report, on a small board without a GPU.

4 stepswith julia-1, laya