OpenAI Agents SDK & Claude 4 Opus IPO Financial Compliance Audit
Late-stage Indian startups preparing for public listing must perform continuous financial reconciliation across Account Aggregator feeds and tax filings. You will build an autonomous compliance agent that reconciles ledger statements, flags tax anomalies, and generates draft IPO filing reports.
What you are building
The core problem, expected build, and operating context for this challenge.
Deploy an OpenAI Agents SDK compliance multi-turn agent utilizing Claude 4 Opus for complex financial schema extraction across Account Aggregator banking streams.
How work is evaluated
Evaluates accuracy in detecting revenue discrepancies across AA streams and trial balances.
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.
Discrepancy Detection Accuracy
Must correctly flag multi-million discrepancy between AA and GST
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Reconciliation Precision
Exactness of calculated revenue variance • target: 1 • range: 0.95-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement OpenAI Agents SDK multi-agent workflows with handoffs and function calling
Leverage Claude 4 Opus reasoning for intricate multi-entity tax and revenue reconciliation
Ingest Account Aggregator financial data feeds in compliant AA payload formats
Generate regulatory DRHP (Draft Red Herring Prospectus) compliance summaries
Reference links and supporting material
Simulated startup financial ledgers, GST return filings, and Account Aggregator bank transactional statements.
Challenge setup
Evaluates accuracy in detecting revenue discrepancies across AA streams and trial balances.
Challenge input
JSON with company financial statements, AA bank feeds, and GST return summary
OpenAI Agents SDK
Manages stateful multi-turn agent handoffs.
Claude 4 Opus
Executes complex financial schema extraction and audit reasoning.
Evaluated output
JSON with audited_revenue, discrepancy_found (bool), and discrepancy_amount
- Must correctly flag multi-million discrepancy between AA and GST
- Exactness of calculated revenue variance • target: 1 • range: 0.95-1
- Benchmark: FinAudit-Bench
- Reconciliation Precision 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
- OpenAI · Optional
- Giskard · Optional
- Claude 4 Opus · Required
Evaluation contract
- Discrepancy Detection Accuracy · Weight 1
- Reconciliation Precision · Weight 1
Metrics
- Reconciliation Precision · Target 1
Benchmarks
- FinAudit-Bench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with company financial statements, AA bank feeds, and GST return summary
Output: JSON with audited_revenue, discrepancy_found (bool), and discrepancy_amount
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Orchestrator
- Claude 4 Opus · Reasoning Engine
Recipe freeze marker
No recipe freeze time is recorded.