FraudTrax

FraudTrax is an analysis engine made for fraud and auto-theft investigation by law enforcement. It reads the evidence in front of you, surfaces what a specialist in each vertical would catch, and ties every flag back to an authority. Fast enough for the curb, accurate enough for the courthouse.

Why we built it

FraudTrax is not a SaaS product dressed up to look friendly to law enforcement. It is a toolkit built from years of running the same temp-tag stops, dealer audits, and theft scenes that officers run every shift. The team came to software engineering from policing for one reason: technology accelerates crime faster than agencies can keep up, and affordable, solid tooling is the missing piece. FraudTrax is the answer to that gap, built by people who have stood on both sides of the problem.

The problem it confronts

The pattern recognition that catches a washed title or a fake temporary tag takes years of cases to build, and it usually lives in one or two veteran investigators per region. When they retire, it walks out the door. Meanwhile the answer an officer needs at the curb (is this VIN consistent, does this dealer lot reconcile, what authority does this violate) often takes days to assemble by hand, if it gets assembled at all. FraudTrax puts that specialist instinct in the officer's hand in seconds, and writes it down in language a court will accept.

How it works

DAN-O reads photos and structured case input and returns articulable indicators. For each one it cites the specific authority it violates (a registry record, a federal advisory, a statute, or a vehicle standard), names what cannot be determined from the evidence, and recommends the next step (NCIC, NMVTIS, NICB, hold). It then drafts a court-ready paragraph the officer can paste into the narrative. The officer signs off. The citation carries through to the printed packet.

The platform covers the two specialties that share evidence, vehicle fraud and auto theft:

• Auto-theft scene documentation captures VIN, damage, forced entry, odometer, and the federal sticker, auto-populates the report from suspect documents, and estimates value.
• VIN verification takes captures across the standard surfaces and flags cross-surface mismatches.
• Plate fraud analyzes temporary tags, paper plates, and dealer-issued plates against state registries and known schemes.
• Dealership audit reconciles lot inventory and books and cites the regulation each violation breaks.
• AI-drafted ROI writes a Report of Investigation in ILETSB format with context-anchored PII redaction (VIN, ORI, badge, and plate kept; DL, SSN, and DOB scrubbed).
• LE-issued forms (VIN certifications, lien releases, title and registration releases) auto-populate from the scan, officer, agency, and suspect data, on agency letterhead.

What makes it different

Every flag references a specific authority, and the output uses articulable language ("exhibits indicators consistent with," not "is stolen"). The printed packet discloses which AI model produced the analysis, which is what admissibility under the Daubert and Frye standards requires. It augments an agency's records management system rather than replacing it, and it runs on a Section 889-clean stack operated by a veteran-owned small business.

Transparency and limits

• DAN-O will not conclude a vehicle is stolen. Only NCIC can. It surfaces indicators and points you to the authority that decides.
• It does not substitute for an officer's judgment, and it does not surface unbacked hunches.
• FraudTrax does not run inside the CJIS perimeter. The architecture assumes any single layer could fail and uses defense in depth: origin gating, tier-based authorization, per-IP and per-account rate limiting, encrypted-at-rest storage, and signature-verified webhooks.