The candidate will design a system where agents parse pipeline specifications, map them to relevant regulatory constraints (e.g., GDPR, data locality policies), and collaboratively reason about compliance violations. The research will involve implementing a prototype that integrates compliance rule sources, pipeline metadata, and multi-agent reasoning strategies. The system will be evaluated on real-world or synthetic data pipelines, assessing accuracy of compliance detection, scalability, and trustworthiness. The contribution will be a novel application of multi-agent LLMs to regulatory and policy assurance in data- intensive systems. Work to be done:
- Literature Review: Survey state-of-the-art in compliance checking for data pipelines, legal AI, and regulatory reasoning with LLMs.
- Requirement Gathering: Identify regulatory frameworks and policies relevant for pipelines (e.g., GDPR, AI Act, data locality rules).
- Knowledge Representation: Formalize a subset of compliance rules (e.g., data minimization, consent, logging) into machine-interpretable formats.
- Agent Design: Define specialized agents such as: o Rule Extraction Agent (maps text to structured constraints)
- Pipeline Analyzer Agent (extracts pipeline metadata and processes)
- Compliance Checker Agent (matches pipeline behavior with rules)
- Reporting Agent (generates compliance reports and recommendations)
- Prototype Development: Implement a working system integrating agents, pipeline metadata, and compliance rules.
- Case Studies: Apply the prototype to synthetic and real-world pipeline examples (e.g., bioinformatics workflows, IoT data pipelines).
- Evaluation: Assess accuracy, false positives/negatives, scalability, and explainability of compliance assessments.
- Discussion: Reflect on trustworthiness, limitations, and future improvements for real-world adoption.