Evaluating LLM Simulators as Differentially Private Data Generators

Build an open-source pipeline that takes DP-protected datasets and maps them to agentic persona simulators. This allows for privacy-preserving synthetic data generation for testing financial workflows.

Suggested repo: persona-privacy-bench

"Generate privacy-compliant synthetic datasets using LLM agents."

Estimated effort: 40h