Contrax is a compliance and audit analysis tool designed to evaluate structured configurations and detect potential risks, inconsistencies, and violations using graph-based modeling and NLP/NLI techniques.
While the current implementation uses AWS services for deployment, the system is not limited to AWS and can be adapted to audit policies, workflows, or rule-based systems across domains.
Contrax ingests structured inputs (such as policies, documents, or configurations), processes them through a pipeline, and constructs a graph representation of relationships and dependencies.
Using this representation, it performs:
- Semantic analysis of rules and conditions
- Detection of inconsistencies or risky patterns
- Identification of unintended interactions between entities
-
Generic Compliance Auditing
Works across domains where rules, policies, or structured constraints exist. -
Graph-Based Analysis
Models entities and relationships to uncover non-obvious dependencies. -
NLP + NLI Integration
Interprets meaning and intent beyond simple rule matching to:- Detect contradictions
- Identify over-permissive or conflicting conditions
- Surface logical inconsistencies
-
Automated Processing Pipeline
Handles ingestion, validation, processing, and result storage.
- Input data (e.g., documents, policies) is uploaded
- Files are validated and queued for processing
- Processing services extract and analyze content
- NLP/NLI models evaluate semantic relationships
- Results are stored and optionally visualized as a graph
- Notifications or outputs are generated for further review
- Compliance auditing for rule-based systems
- Policy validation and consistency checking
- Risk analysis in structured configurations
- Python
- NLP / NLI models
- Graph-based data modeling using Neo4j
- AWS (S3, Lambda, SQS, SNS) for pipeline orchestration
- React Js
- The current deployment leverages AWS for scalability and event-driven processing
- The core design is platform-agnostic and can be extended beyond AWS environments