Skip to main content
Back to challenges
Challenge

AutoGen MSME Credit Risk Underwriting Agent

Non-banking financial institutions (NBFCs) require automated multi-agent risk assessment for underwriting under-served MSME loans. You will create a multi-agent conversational loop that parses Account Aggregator statements, GST returns, and field audit notes to issue automated credit recommendations.

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.

Build an AutoGen multi-agent system where auditor, tax analyst, and risk assessment agents collaboratively evaluate MSME creditworthiness from DPI data sources.

Delivery guide

How work is evaluated

Evaluation

Evaluates multi-agent conversation consensus and credit limit generation accuracy.

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

Over-Leverage Protection Test

Must reject loans exceeding 3x annual GST turnover

Binary check

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

Dimension 2

Consensus Quality Score

Rigor of multi-agent validation discussion steps • target: 0.95 • range: 0.85-1

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 conversational multi-agent groups using AutoGen AssistantAgent and UserProxyAgent

  • Implement structured dialogue loops between specialized financial roles

  • Process Account Aggregator banking transaction trends for cashflow verification

  • Compute automated credit score bounds and sanction limits for small businesses

Resources and assets

Reference links and supporting material

Dataset notes

Anonymized profile records of MSMEs containing bank balance trends, GST filings, and repayment histories.

Challenge setup

Evaluates multi-agent conversation consensus and credit limit generation accuracy.

Challenge input

JSON with MSME profile, monthly avg balance, GST turnover, and requested loan

AutoGen

Provides multi-agent conversational orchestration.

Evaluated output

JSON with decision (APPROVED|REJECTED), sanctioned_amount, and interest_rate

Evaluation checks
  • Must reject loans exceeding 3x annual GST turnover
  • Rigor of multi-agent validation discussion steps • target: 0.95 • range: 0.85-1
Configured evaluation evidence
  • Benchmark: CreditUnderwriting-Bench
  • Consensus Quality Score target: 0.95
  • 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
  • Resemble AI · Optional
Observation
  • Gentrace · Optional
Reward / Eval
  • Gentrace · Optional

Evaluation contract

  • Over-Leverage Protection Test · Weight 1
  • Consensus Quality Score · Weight 1

Metrics

  • Consensus Quality Score · Target 0.95

Benchmarks

  • CreditUnderwriting-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with MSME profile, monthly avg balance, GST turnover, and requested loan

    Output: JSON with decision (APPROVED|REJECTED), sanctioned_amount, and interest_rate

    Execution harness: Python

Evaluation tools

  • AutoGen · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about AutoGen MSME Credit Risk Underwriting Agent