Skip to main content
Back to challenges
Challenge

Ethics Compliance Agent for Financial Market Analysis

Develop a cutting-edge multi-agent system using Microsoft AutoGen to perform real-time ethical and regulatory compliance analysis for financial market activities. Inspired by recent headlines regarding market integrity rules, this system will leverage a team of specialized agents to interpret complex market news, cross-reference legal frameworks (such as insider trading regulations), and flag potential violations. Qwen 3 235B will power the core reasoning capabilities for the agents, enabling nuanced understanding of verbose financial texts and intricate regulatory guidelines. The solution emphasizes modern AI interaction and evaluation. It should integrate seamlessly with Fixie for natural language voice interactions, allowing financial stakeholders to query compliance statuses and receive immediate alerts or summaries through conversational interfaces. Patronus AI will be employed for robust evaluation and monitoring of the agents' behavior, ensuring strict adherence to compliance policies, identifying any biases, or detecting errors in their decision-making processes. Furthermore, a Wix dashboard will be created to visualize compliance reports, agent decision-making rationales, and overall system performance, providing transparency and auditability.

Workflow AutomationHosted 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 powered by Qwen 3 235B to ensure ethical compliance in financial market analysis with voice interaction and AI-driven evaluation.

Delivery guide

How work is evaluated

Evaluation

Evaluate the multi-agent system's ability to accurately identify compliance issues, its reasoning transparency, and its responsiveness via the Fixie voice interface.

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
6 scoring checks
Binary
6 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

CorrectViolationIdentification

The system correctly identifies the primary compliance violation (e.g., 'Insider Trading') when present.

Binary check

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

Dimension 2

ReasoningClarity

The 'reasoning' provided by the agent system is clear, concise, and directly supports the compliance status.

Binary check

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

Dimension 3

NoFalsePositives

The system does not flag violations for scenarios that are clearly compliant according to the rules.

Binary check

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

Dimension 4

Compliance Accuracy

The percentage of test cases where the 'compliance_status' matches the ground truth. • target: 0.95 • range: 0-1

Binary check

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

Dimension 5

Reasoning Quality Score

A subjective score (1-5) based on the coherence, completeness, and correctness of the agent's reasoning. • target: 4 • range: 1-5

Binary check

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

Dimension 6

Voice Response Latency (seconds)

The average time taken for the system to process a voice query via Fixie and return a spoken response. • target: 2 • range: 0-5

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

  • Master Microsoft AutoGen for orchestrating complex, conversational multi-agent workflows, defining agent roles and communication patterns.

  • Implement ethical guidelines and regulatory compliance checks within an AI agent system, focusing on financial market rules like insider trading prohibitions.

  • Integrate Qwen 3 235B as the backbone LLM for advanced text comprehension, legal document analysis, and sophisticated reasoning within agents.

  • Design voice-enabled interfaces using Fixie, allowing stakeholders to interact with the compliance system through natural language queries and receive spoken responses.

  • Utilize Patronus AI for robust evaluation and monitoring of agent behavior, ensuring compliance adherence, identifying potential biases, and tracing decision rationales.

  • Develop custom tool integrations for AutoGen agents, enabling them to access mock financial databases or legal rule engines for information retrieval.

  • Build a dynamic dashboard using Wix developer tools to visualize compliance reports, agent interactions, and audit trails.

Resources and assets

Reference links and supporting material

Dataset notes

Sample data for 1 tasks

Challenge setup

Evaluate the multi-agent system's ability to accurately identify compliance issues, its reasoning transparency, and its responsiveness via the Fixie voice interface.

Challenge input

JSON object with 'market_news' (string), 'trade_data' (JSON object with 'user', 'stock', 'action', 'volume', 'time'), and 'ethical_rules' (array of...

AutoGen

Agent Systems · Multi-Agent Systems

Patronus AI

Evaluation and guardrail platform.

Qwen 3

Policy Serving in the agent workflow.

Evaluated output

JSON object with 'compliance_status' (string), 'reasoning' (string), 'flagged_violations' (array of strings), and 'confidence' (float 0.0-1.0).

Evaluation checks
  • The system correctly identifies the primary compliance violation (e.g., 'Insider Trading') when present.
  • The 'reasoning' provided by the agent system is clear, concise, and directly supports the compliance status.
  • The system does not flag violations for scenarios that are clearly compliant according to the rules.
Configured evaluation evidence
  • Compliance Accuracy target: 0.95
  • 1 public reference case
Configured runtime
  • 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

Action Space
  • AutoGen · Required
  • Patronus AI · Optional
Policy Serving
  • Qwen 3 · Optional

Evaluation contract

  • CorrectViolationIdentification · Weight 1
  • ReasoningClarity · Weight 1
  • NoFalsePositives · Weight 1
  • Compliance Accuracy · Weight 1
  • Reasoning Quality Score · Weight 1
  • Voice Response Latency (seconds) · Weight 1

Metrics

  • Compliance Accuracy · Target 0.95
  • Reasoning Quality Score · Target 4
  • Voice Response Latency (seconds) · Target 2

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON object with 'market_news' (string), 'trade_data' (JSON object with 'user', 'stock', 'action', 'volume', 'time'), and 'ethical_rules' (array of strings).

    Output: JSON object with 'compliance_status' (string), 'reasoning' (string), 'flagged_violations' (array of strings), and 'confidence' (float 0.0-1.0).

    Execution harness: Python

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Ethics Compliance Agent for Financial Market Analysis