Geodineum Business · Anomaly

The one record
that does not belong.

You have the data. GeoV finds the point that breaks the pattern, and tells you why.

Learned normal, dashed. Flagged outliers, red.

What it does

Distance from normal, scored.

GeoV learns the shape of normal from data you trust. Every new record gets a distance and a percentile. Not pass or fail. A number you can rank, threshold, and act on.

It tells you why

Glass box, not black box.

A score you cannot explain is a dead end. GeoV breaks each one down by field, and by field pair. You see which columns raised the alarm, and which combinations look wrong together.

No rules to write

Correlations, learned.

Fraud and defects hide between fields, not in any single value. GeoV learns those correlations from the covariance of your data. Change the schema, keep the pipeline.

Deterministic

Same input, same score.

The scoring path runs on gMath fixed-point arithmetic. Designed for reproducible, auditable results. Verified on x86-64.

Where it fits

Point it at trusted history. Score the new stuff. Finance and fraud, cybersecurity telemetry, manufacturing quality, any stream where the signal is the rare record that deviates.

Work with us

Score your data.

Bring a trusted dataset and a schema. We learn its geometry and let the outliers surface themselves. On-premises for regulated or sensitive data, managed for everything else.

The first request we take on is free. Conditions apply.

Sensitive or regulated data is handled on-premises, in your own environment. Everything else can run on our managed infrastructure.