Build a Fiat-Crypto Banking Access Evaluator using OpenAI Agents SDK and Claude 4.1 Opus
Crypto businesses face persistent hurdles maintaining correspondent banking relationships due to complex risk profiling demands. Build a risk evaluation agent using OpenAI Agents SDK with Claude 4.1 Opus to analyze crypto exchange financial balance sheets and compliance procedures. Calculate risk scores matching expert underwriter standards with 95% alignment.
What you are building
The core problem, expected build, and operating context for this challenge.
Deploy an automated banking risk evaluator using OpenAI Agents SDK and Claude 4.1 Opus to score crypto asset service providers for bank account eligibility.
How work is evaluated
Evaluate the risk scoring consistency and decision accuracy for prospective crypto corporate accounts.
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.
Reject Insolvent Firms
Ensure firms with liabilities exceeding reserves are flagged REJECT.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Underwriter Score Correlation
Pearson correlation coefficient between agent tier_score and expert human risk ratings. • target: 0.95 • range: -1-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Utilize OpenAI Agents SDK to manage multi-turn risk underwriting interviews
Integrate Claude 4.1 Opus for in-depth analysis of financial statements and audit disclosures
Produce compliant risk rating reports for bank onboarding committees
Identify high-risk liquidity dependencies and unbacked asset liabilities
Reference links and supporting material
Synthetic dataset of 500 financial profiles for digital asset brokers and crypto platforms seeking fiat bank account services.
Challenge setup
Evaluate the risk scoring consistency and decision accuracy for prospective crypto corporate accounts.
Challenge input
JSON containing firm_name, asset_reserves_usd, liabilities_usd, aml_audit_score, and operating_jurisdictions.
OpenAI Agents SDK
Provides function execution loops and state management for multi-step reasoning.
Claude 4.1 Opus
High performance on complex document understanding and formal risk modeling.
Evaluated output
JSON containing risk_category ('LOW', 'MEDIUM', 'HIGH', 'REJECT'), tier_score (0-100), and rationale.
- Ensure firms with liabilities exceeding reserves are flagged REJECT.
- Pearson correlation coefficient between agent tier_score and expert human risk ratings. • target: 0.95 • ra...
- Benchmark: BankRiskEval
- Underwriter Score Correlation target: 0.95
- 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
- OpenAI · Required
- Claude Code · Optional
- Claude Code · Optional
- Claude 4.1 Opus · Optional
Evaluation contract
- Reject Insolvent Firms · Weight 1
- Underwriter Score Correlation · Weight 1
Metrics
- Underwriter Score Correlation · Target 0.95
Benchmarks
- BankRiskEval
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing firm_name, asset_reserves_usd, liabilities_usd, aml_audit_score, and operating_jurisdictions.
Output: JSON containing risk_category ('LOW', 'MEDIUM', 'HIGH', 'REJECT'), tier_score (0-100), and rationale.
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Architecture
- Claude 4.1 Opus · Core LLM Engine
Recipe freeze marker
No recipe freeze time is recorded.