VLM · AWS-native ETL · insurance / fleet

AutoClaim-VLM

A vehicle-damage assessment pipeline: raw claim photo in, schema-valid damage JSON out — routed by severity and confidence to repair workflows, ERP, or a human adjuster.

PaliGemma 3B VLM Glue → SageMaker → Redshift confidence routing Terraform IaC
The problem

Adjuster review is slow and subjective

The traditional human adjuster cycle for a damage claim runs 2–5 days and $45–$120 per assessment, with inconsistent descriptions and manual routing.

The goal: a sub-90-second automated path that emits typed damage records — class, severity, zones, repair estimate, routing action — directly consumable by repair workflows, SAP ERP and claims platforms, while feeding a warehouse for trend analytics.

The pipeline

Ingest → transform → classify → load → serve

Mobile / kiosk / IoT camera │ ▼ S3 (WORM) ─► Lambda ─► SQS FIFO ─► Glue (PySpark) ─► SageMaker VLM (PaliGemma 3B) │ │ blur / pHash / 640×640 schema-valid damage JSON │ │ ▼ ▼ Redshift + DynamoDB ◄── confidence-based routing │ (auto-approve · repair · adjuster) ▼ QuickSight BI + SNS alerts + SAP hook
Transform

Quality-gated ETL

Glue PySpark: Laplacian blur scoring, perceptual-hash dedup, 640×640 normalize, Great Expectations checks before any write.

Classify

VLM + schema

SageMaker-hosted VLM; every output validated against a draft-07 JSON Schema — invalid responses caught and escalated, never silently written.

Route

Confidence tiers

Severity × confidence → auto-approve, queue-for-repair, flag-review or route-to-adjuster, with total-loss override.

Design targets

Sized for, not yet benchmarked

<45s
p50 latency (target)
>94%
accuracy (target)
~$0.004
per image (target)
~160k
corpus images

Open-source ingestion + label normalization across five datasets (CompCars, CarDD, Stanford Cars, Kaggle, HuggingFace) is implemented; the PaliGemma 3B fine-tune is the next milestone.

Stack

Built with

Python 3.11AWS Glue (PySpark)SageMaker PaliGemma 3BRedshiftDynamoDB SQS FIFOStep FunctionsGreat Expectations TerraformCloudWatch + X-Ray