InsuranceAI Services

AI Insurance Damage Assessment

Vehicle damage detection, claim workflow and fraud checks for a US InsurTech company. Team of 7.

Key Details

Challenge	A US InsurTech company needed to assess vehicle damage, move claims through a workflow and flag fraud from the same platform.
Solution	Computer-vision damage and severity detection, claim workflow automation, fraud detection and dashboards.
Technologies	YOLO, TensorFlow, OpenCV, Python, FastAPI, React, AWS, PostgreSQL, Amazon S3, Grafana, Docker, REST

Technologies used

YOLOTensorFlowOpenCVFastAPIReactAWSPostgreSQLAmazon S3GrafanaDocker

Client

The client is a fast-growing InsurTech SaaS company in the United States. It provides digital claim processing for insurance providers and automotive partners, and wanted AI assessment of vehicle photos in the claim workflow.

Product screenshots

Light claims web app showing a photo of a silver sedan with labelled outlines over a door dent, door scratches, a front fender scratch and front bumper scuffs, a severity legend, and a draft repair estimate table listing part, damage type, suggested action and confidence with accept and edit controls. — Adjuster view: the model outlines each damaged area on the claim photo, names the part and damage type, and drafts a repair line for the adjuster to confirm.

Two phones on a mint background: the first shows a camera capture screen with a car outline guide, an angle checklist and a shutter button; the second shows the claim tracker with a progress timeline, the uploaded photos, and a message that an adjuster is reviewing the AI assessment. — Policyholder app: a guided capture flow walks the driver around the car one angle at a time, then shows the claim status as it moves through review.

What we built

Models that detect scratches, dents, broken parts and impact areas from uploaded photos, and classify severity
Claim intake, image validation and damage-assessment summaries
Workflow integration with claim management systems
Fraud checks for duplicate or suspicious images and abnormal claim patterns
Dashboards for claim tracking and AI analysis

Engagement

Delivery is time and materials, run as Agile/Scrum, with 7 engineers over 10 months and is ongoing.

Project team: 7 engineers. Duration: 10 months, ongoing. Model: Time & materials, Agile/Scrum.

Services used

AI Services →Workflow AI Agent →