Skip to main content
Back to challenges
Challenge

Account Aggregator Credit Underwriting Multi-Agent System using AutoGen

Consent-based credit evaluation across Account Aggregator financial data requires structured multi-agent verification. Build a multi-agent financial auditing workflow using AutoGen and Pydantic AI data validation to automate 100 mock loan applications with zero PII leakage and 98% rule compliance.

Business OperationsHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Construct an automated financial auditing multi-agent workflow using Microsoft AutoGen and Pydantic AI for structured schema validation.

Delivery guide

How work is evaluated

Evaluation

Evaluates multi-agent conversation history and final JSON loan decision payload.

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

pii_leakage_check

Ensures raw unmasked PAN or account numbers do not leak in conversation logs.

Binary check

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

Dimension 2

rule_compliance

Percentage of correct underwriting recommendations based on strict debt-to-income ratio rules. • target: 0.98 • 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

  • Build multi-agent conversational patterns using AutoGen

  • Incorporate Pydantic AI for strict runtime data validation and output parsing

  • Process Account Aggregator (AA) structured financial statements

  • Verify credit scoring rules against automated compliance logic

Resources and assets

Reference links and supporting material

Dataset notes

100 synthetic Account Aggregator applicant profiles containing transaction history and credit scores.

Challenge setup

Evaluates multi-agent conversation history and final JSON loan decision payload.

Challenge input

JSON object with applicant_id, monthly_income, debt_obligations, and aa_bank_statement array.

AutoGen

Facilitates multi-agent conversation and delegation.

Pydantic AI

Ensures output compliance and structural verification.

Evaluated output

JSON object with recommendation ('APPROVED', 'REJECTED', 'MANUAL_REVIEW'), max_eligible_loan, and compliance_score.

Evaluation checks
  • Ensures raw unmasked PAN or account numbers do not leak in conversation logs.
  • Percentage of correct underwriting recommendations based on strict debt-to-income ratio rules. • target: 0....
Configured evaluation evidence
  • Benchmark: Fintech Rule Engine Benchmark
  • Rule Compliance target: 0.98
  • 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
  • AutoGen · Required
  • Pydantic AI · Optional
Observation
  • Gentrace · Optional
Reward / Eval
  • Gentrace · Optional
Orchestration
  • Pydantic AI · Optional

Evaluation contract

  • pii_leakage_check · Weight 1
  • rule_compliance · Weight 1

Metrics

  • rule_compliance · Target 0.98

Benchmarks

  • Fintech Rule Engine Benchmark
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON object with applicant_id, monthly_income, debt_obligations, and aa_bank_statement array.

    Output: JSON object with recommendation ('APPROVED', 'REJECTED', 'MANUAL_REVIEW'), max_eligible_loan, and compliance_score.

    Execution harness: Python

Evaluation tools

  • AutoGen · Multi-Agent Orchestrator
  • Pydantic AI · Data Validation

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Account Aggregator Credit Underwriting Multi-Agent System using AutoGen