Skip to main content
Back to challenges
Challenge

Post-Hurricane Claims Fraud Detection with Pydantic AI and Qwen 3

Following major natural disasters, fraudulent property claims surge, leading to multi-million dollar losses. You will build a type-safe claims fraud analysis pipeline using Pydantic AI and Qwen 3 to evaluate damage reports against satellite imagery data. Achieve a fraud classification F1-score of 0.88 or higher on hurricane claim benchmarks.

Machine LearningHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Integrate Pydantic AI schemas with Qwen 3 vision-language inference to validate post-disaster insurance claims.

Delivery guide

How work is evaluated

Evaluation

Evaluate claim validation accuracy and fraud classification metrics against labeled test claims.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

schema_conformance

Output strictly adheres to required FraudAssessment Pydantic model.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

f1_score

F1 score for binary fraud classification • target: 0.88 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Construct strictly typed claim validation models using Pydantic AI

  • Process multimodal imagery and text claims data using Qwen 3

  • Evaluate anomaly scores across claim histories and physical damage metrics

  • Deploy a structured fraud detection pipeline with reproducible evaluation metrics

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic and anonymized post-hurricane insurance claim dataset containing text summaries, imagery metadata, and ground truth fraud labels.

Challenge setup

Evaluate claim validation accuracy and fraud classification metrics against labeled test claims.

Challenge input

JSON containing claim text, damage description, and image URL

Pydantic AI

Structured outputs and type-safe agent orchestrations.

Qwen 3

Multimodal reasoning over document text and claim images.

Evaluated output

JSON object matching FraudAssessment Pydantic schema

Evaluation checks
  • Output strictly adheres to required FraudAssessment Pydantic model.
  • F1 score for binary fraud classification • target: 0.88 • range: 0-1
Configured evaluation evidence
  • Benchmark: ClaimsFraudBench-2026
  • F1 Score target: 0.88
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Pydantic AI · Optional
  • Aim · Optional
Policy Serving
  • Qwen 3 · Required
Orchestration
  • Pydantic AI · Optional

Evaluation contract

  • schema_conformance · Weight 1
  • f1_score · Weight 1

Metrics

  • f1_score · Target 0.88

Benchmarks

  • ClaimsFraudBench-2026
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing claim text, damage description, and image URL

    Output: JSON object matching FraudAssessment Pydantic schema

    Execution harness: Python

Evaluation tools

  • Pydantic AI · Agent Framework
  • Qwen 3 · LLM Provider

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Post-Hurricane Claims Fraud Detection with Pydantic AI and Qwen 3