Regression and ML-based scoring and anomaly detection, computed inside the live payment pipeline.
Scoring and risk models usually run in a separate environment, reached from the payment stream across an integration boundary. That boundary adds latency and operational surface, and it widens the window in which a fraudulent payment can clear.
Streaming tools such as Flink move and filter the payment data, but the model and the decision run elsewhere, after the data has crossed into another system.
In contrast, SpaceCell Lightning runs sub-millisecond, and enables you to embed any Python-based ML models directly in the call path, leaving you with more time and power to run comprehensive scenarios before the authorisation window.
Lightning runs regression-based scoring and anomaly detection inside the live payment pipeline. A transaction arrives, the model scores it against its rolling behavioural window, and the decision executes in one engine.
The full statistical suite runs on the stream, with a Python bridge for the machine-learning models your risk team already maintains.
Each payment is scored against its rolling behavioural window as it arrives.
Unusual frequency, amount, and pattern shifts are detected on the live stream.
Continuous monitoring runs against thresholds, with breaches routed for review.
High-risk transactions are flagged and held before settlement.
Lightning runs the live scoring and decision logic for the payment path - the statistical models on the stream, and a Python bridge for the machine-learning models your risk team already maintains.
It is the scoring and decision engine, not a case-management suite, a payments gateway, or a model-training platform. It computes the decision and routes it to the systems that act on it.
No. Lightning is the live scoring and decision engine. It runs the statistical models on the payment stream and routes the decision. Your case management, investigation, and rules platform stay in place.
Lightning is written in Rust and calls any standard REST API or integration endpoint, so it runs beside your rails and fraud systems and routes its decisions to them. If you are unsure how it maps to your stack, reach out and we can help you scope it.
It is a single-process executable binary with a very small, auditable set of dependencies, largely built in-house. We provide deployable software only: you install and operate it entirely within your own environment. We do not host, operate, access, receive, store, transmit, log, or inspect the cardholder, transaction, or personal data it processes. Because we never handle your data, SpaceCell is not a data processor under UK GDPR, and no data processing agreement is required for this model. PCI DSS and data-protection responsibilities for running the software remain with you.
No. You keep full control of your data, and the software runs inside your own environment. To analyse live transactions, the feed becomes an input to Lightning, which processes it in place for immediate insight and action. The data never leaves your walls.