ComplianceAI

Policy Parsing: Sample policy PDFs are processed using pdfplumber to extract text efficiently.

Rule Interpretation: Extracted text is converted into structured JSON rules using Gemini LLM, following a predefined schema.

Rule Execution Engine: The Python-based rule engine interprets JSON rules and applies them to the IBM AML dataset.

• Supports threshold, frequency, and payment method rules.

• Violations are flagged with rule IDs for clear audit trails.

Robust & Safe: Rules with missing fields are handled gracefully, ensuring consistent execution.

Metrics Logic: Precision, recall, and F1-score calculations are built into the engine for evaluation.

Demo Output: Prototype successfully identifies policy violations, demonstrating the system’s core capabilities.