EHRSynth is a synthetic data studio for clinical datasets. Upload any EHR-style CSV, and it infers column types and care stages (baseline, pre-op, intra-op, post-op, follow-up), removes identifiers, trains a stage-conditional diffusion model and generates synthetic patients you can download.

Each synthetic dataset is evaluated for fidelity (Wasserstein and total variation distance), privacy (membership inference, replica rate, attacker AUC) and utility (train on synthetic, test on real).