Vehicle Intelligence
A mid-sized LPR deployment generates millions of reads per day from fixed cameras, mobile units, and partner networks, each vendor with its own schema, its own confidence scoring, and its own idea of what a timestamp means.
Mentat builds the layer that turns that into one coherent, queryable, governed dataset. Reads are normalized on ingest, deduplicated across overlapping coverage, geo-resolved, and written into storage that can answer a query in seconds and delete itself on schedule.
Equally important is what the platform refuses to do. Purpose limitation, retention ceilings, and query justification are enforced by the system, which is what turns an LPR program from a liability into a defensible capability.
What we build
A pipeline built for sustained high-cardinality write load, with backpressure handling so a network partition delays reads instead of losing them.
Raw reads become comparable records: consistent plate formatting, resolved jurisdictions, calibrated confidence, and honest handling of ambiguity.
Retention is not a setting someone remembers to check. It is executable policy, with different clocks for general reads and reads attached to an active matter.
Every search is attributed to a person, a purpose, and a time, producing the record that internal affairs, an auditor, or a court will ask to see.
How we get there
Every fixed site, mobile unit, and partner feed catalogued with its vendor, schema, coverage area, and data-sharing terms.
One canonical read format, applied at the boundary, so downstream systems never have to know which vendor produced a record.
Retention windows, purpose codes, access tiers, and sharing agreements expressed as enforced configuration rather than written guidance.
Query performance, camera health, deletion confirmations, and access patterns monitored continuously and reportable on demand.
Governance & safeguards
LPR is among the most scrutinized technologies an agency can deploy. Retrofitting privacy controls onto a system designed without them is expensive and rarely convincing. We build them into the data model from the first sprint.
Expiry produces a verifiable record that specific data was removed on schedule, across primary storage, replicas, and backups.
There is no query path that bypasses attribution. Service accounts, integrations, and bulk exports carry the same accountability as a console search.
Partner and regional network sharing is governed by explicit, versioned agreements the platform enforces, including what may be shared and for how long.
Built on
We build on established open-source and cloud-native building blocks rather than proprietary black boxes, so the platform stays inspectable, portable, and yours.
Ingest
Storage & query
Controls
Related
Running or planning an LPR program?
We can review your ingest path, retention posture, and audit coverage, then show you what a governed LPR platform looks like at your read volume.