CarInfo helps specialty-vehicle lenders reconcile seller claims, inspections and appraisals before funding. Start with a listing. Get a source-linked case and the evidence still needed for a human review.
Try the underwriting deskTry a fictional case without signing in, or research a live listing. Built for human review; no automated premiums or credit decisions. AI review is available only to enabled pilot accounts after provider activation.
The reasoning stays connected to the evidence that existed at the time.
Start with a listing link or VIN. Retrieve available listing metadata and NHTSA records. Source restrictions, missing facts and model-only findings remain visible.
Keep seller claims, official records and reviewer-accepted documents distinct. The AI review integration compares the case evidence and returns cited suggestions; a human must assess those suggestions.
Use separate collateral and insurance views. Explicit rules identify evidence gaps. Save the reviewer’s rationale alongside a dated assessment and its rule version.
Attach documented repair expenses, losses or collateral sale proceeds later. Outcomes stay private. No cross-customer training or proprietary loss dataset is claimed.
The opportunity we’re testing: make the evidence easier to assemble, reconcile and review.
J.J. Best states that a loan will not fund until its collateral verification is completed. Its FAQ also identifies modified vehicles among the cars that can require an appraisal.
J.J. Best · Vehicle verification and appraisal FAQWoodside Credit describes vehicle verifications and appraisals for some transactions, especially private-party purchases, and already uses virtual inspections. CarInfo’s proposed role is reconciling the resulting evidence, not replacing those inspections.
Woodside Credit · Inspection and appraisal FAQASOP 56 addresses a model’s intended use, data and assumptions, testing, output validation and documentation. Our current rules identify review requirements; they do not estimate claims frequency or severity.
Actuarial Standards Board · ASOP 56The NAIC’s model bulletin sets expectations for insurers’ AI governance and accurate, lawful consumer decisions. Its implementation depends on jurisdiction. CarInfo’s pilot keeps source references and human review visible; it is not a compliance certification.
NAIC · Insurer AI model bulletinThe OCC’s installment-lending guidance discusses collateral coverage alongside repayment capacity and other underwriting factors. CarInfo’s collateral view examines the vehicle and the selected value basis; it does not assess the borrower.
OCC · Installment Lending handbookThese firms are research sources, not customers or partners. Their requirements establish that the workflow exists; they do not prove demand for CarInfo. A pilot must measure review time, unresolved evidence and reviewer agreement.
Established providers already supply VIN-based insurance rating data. Our first focus is the reviewer handling modified and specialty-vehicle collateral: where installed equipment, condition and supported value need more context than a factory specification.
The asset to develop is permissioned, dated evidence connected to review decisions and actual outcomes. The long-term aim is to learn which documented discrepancies matter to real outcomes. That requires explicit data rights, enough observations and independent validation; it is not a capability this pilot already claims.
Built by Harsh Kapil Vyas, drawing on nine to ten years of involvement in the automotive space. This is a working pilot, not a claim of actuarial credentials or institutional adoption.
Existing market · Verisk VIN-based rating APINo. It organizes vehicle evidence, applies explicit review rules and supports cited AI review when enabled. Loss pricing would require authorized claims and exposure data, suitable methods, independent validation and the relevant institutional and jurisdictional controls.
No. A missing record is not a measured loss probability. The collateral ratio uses a valuation selected and accepted by a reviewer. The app does not approve loans, deny coverage or quote premiums.
Time to assemble and review a file, important unsupported claims identified, evidence requests resolved and reviewer agreement. Later outcomes remain linked to the original evidence. No improvement or proprietary predictive advantage is claimed before measurement.