Orchestrate an Account Aggregator NBFC Credit Underwriting Crew using CrewAI and Synthflow
With major global institutions taking large stakes in Indian NBFC arms, rapid and compliant digital underwriting via India's Account Aggregator (AA) framework is critical. In this challenge, developers will build an automated multi-agent credit risk assessment system using CrewAI, integrated with Synthflow to handle conversational loan intake interviews. The crew consists of specialized agents: an AA Data Parser Agent, a Merchant Financial Risk Evaluator Agent, and a Regulatory Compliance Auditor Agent.
What you are building
The core problem, expected build, and operating context for this challenge.
Design a CrewAI multi-agent crew paired with Synthflow conversational intake to automate NBFC credit risk scoring and Account Aggregator consent ledger auditing.
How work is evaluated
Evaluates the multi-agent crew's output memo for credit decision accuracy, risk classification, and statutory policy adherence.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
credit_decision_alignment
Decision must accurately reflect solvency rules without missing required risk metrics.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
memo_completeness
Score measuring presence of key credit indicators and compliance checks. • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Construct a CrewAI multi-agent hierarchy with clear Agent roles, goals, and backstories.
Integrate Synthflow webhook events for conversational voice/text loan application intake.
Process Account Aggregator (AA) financial consent payloads to derive GST returns, cashflow volatility, and debt-to-income ratios.
Ensure SEBI and RBI regulatory compliance checks are executed autonomously within the crew workflow.
Reference links and supporting material
Dataset containing 100 simulated NBFC SME loan applications complete with Synthflow conversational intake summaries and Account Aggregator FIU financial statement extracts.
Challenge setup
Evaluates the multi-agent crew's output memo for credit decision accuracy, risk classification, and statutory policy adherence.
Challenge input
JSON containing Synthflow transcript and AA statement summary
CrewAI
Orchestrates multi-agent role-playing credit evaluation workflow.
Synthflow
Handles voice/conversational intake for loan applicants.
Evaluated output
JSON formatted CreditMemo containing decision, max_sanction_amount, risk_grade, and compliance_pass
- Decision must accurately reflect solvency rules without missing required risk metrics.
- Score measuring presence of key credit indicators and compliance checks. • target: 1 • range: 0-1
- Benchmark: Fintech-Credit-Bench
- Memo Completeness target: 1
- 1 public reference case
- 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
- Synthflow · Required
- CrewAI · Optional
- Zed · Optional
Evaluation contract
- credit_decision_alignment · Weight 1
- memo_completeness · Weight 1
Metrics
- memo_completeness · Target 1
Benchmarks
- Fintech-Credit-Bench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing Synthflow transcript and AA statement summary
Output: JSON formatted CreditMemo containing decision, max_sanction_amount, risk_grade, and compliance_pass
Execution harness: Python
Evaluation tools
- CrewAI · Agent Framework
- Synthflow · Assistants & Interfaces
Recipe freeze marker
No recipe freeze time is recorded.