Multi-Modal Research Dataset Normalization Pipeline

Design a comprehensive data normalization strategy for heterogeneous scientific research datasets collected across multiple experimental modalities. Create a modular Python workflow that can handle disparate data formats from microscopy, spectroscopy, genomic sequencing, and sensor measurements. The pipeline must include automated data type inference, missing value imputation strategies specific to scientific domains, outlier detection using robust statistical methods, and generate comprehensive metadata logs documenting all transformations. Include error handling for edge cases and provide a flexible configuration mechanism allowing domain researchers to customize preprocessing parameters.