Secure Enterprise Financial Automation
Develop an autonomous agent system using OpenAI Agents SDK to automate complex financial operations within an enterprise setting. This challenge requires building a multi-agent orchestration layer capable of interacting with various financial data sources and enterprise APIs securely. The system must demonstrate reliable execution of tasks such as client vetting, transaction processing, or trade automation, while ensuring strict compliance and audibility. Performance and reliability will be evaluated using adaptive experimentation principles.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a secure, enterprise-grade multi-agent system using OpenAI Agents SDK to automate financial workflows, integrating with external tools and evaluated with adaptive experimentation.
How work is evaluated
The evaluation will focus on the agent system's ability to accurately and securely execute a series of simulated financial tasks, adhere to compliance rules, and demonstrate robustness under varying conditions. Ax will be used to track success rates and decision quality.
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.
CorrectStatus
Output 'status' matches expected outcome.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
NoComplianceViolations
No compliance flags raised for compliant inputs.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
SuccessfulToolUse
Evidence of appropriate tool/API calls in agent trace.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Accuracy
Percentage of tasks executed correctly. • target: 90 • range: 0-100
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Latency (ms)
Average time taken for a task completion (for Groq-optimized parts). • target: 500 • range: 0-5000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master the OpenAI Agents SDK for building complex, stateful agentic workflows including function calling and tool use.
Implement secure authentication and authorization mechanisms for agent access to sensitive financial enterprise APIs.
Design a robust evaluation harness using Ax (Adaptive Experimentation) to continuously test and optimize agent reliability and accuracy in various financial scenarios.
Integrate Groq Cloud to accelerate inference for specific, high-throughput agent tools or sub-agents requiring low-latency responses, potentially for real-time market data analysis.
Build a persistent knowledge base for compliance documents and client histories using Weaviate, enabling advanced RAG capabilities for agents to make informed decisions.
Orchestrate collaborative agent teams (e.g., a 'Vetting Agent' working with a 'Transaction Agent') using OpenAI Agents SDK's capabilities for structured communication and task hand-off.
Develop error handling and recovery strategies for agents operating in a high-stakes financial environment.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The evaluation will focus on the agent system's ability to accurately and securely execute a series of simulated financial tasks, adhere to compliance rules, and demonstrate robustness under varying conditions. Ax wil...
Challenge input
{'client_id': 'string', 'name': 'string', 'address': 'string', 'employment': 'string', 'financial_history': 'list'}
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'status': 'string', 'compliance_flags': 'list', 'summary': 'string'}
- Output 'status' matches expected outcome.
- No compliance flags raised for compliant inputs.
- Evidence of appropriate tool/API calls in agent trace.
- Accuracy target: 90
- 2 public reference cases
- Python execution harness
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
No tool records are attached.
Evaluation contract
- CorrectStatus · Weight 1
- NoComplianceViolations · Weight 1
- SuccessfulToolUse · Weight 1
- Accuracy · Weight 1
- Latency (ms) · Weight 1
Metrics
- Accuracy · Target 90
- Latency (ms) · Target 500
Reference cases
- 2 public reference cases
Workflow tasks
- Task 1
Input: {'client_id': 'string', 'name': 'string', 'address': 'string', 'employment': 'string', 'financial_history': 'list'}
Output: {'status': 'string', 'compliance_flags': 'list', 'summary': 'string'}
Execution harness: Python
- Task 2
Input: {'trade_id': 'string', 'symbol': 'string', 'quantity': 'integer', 'max_price': 'float', 'risk_tolerance': 'string'}
Output: {'status': 'string', 'executed_price': 'float', 'message': 'string'}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.