A GenAI Pipeline for Violinist Kinematic Data Management

Health data management remains a major challenge in digital health research, particularly for complex biomechanical datasets requiring cleaning, harmonization, and preparation for downstream analysis. This paper presents a Generative Artificial Intelligence (GenAI)-supported pipeline for managing an upper-body kinematic dataset from 26 violinists with and without recent neck-shoulder pain. The workflow was designed to support dataset profiling, structured cleaning, schema harmonization, metadata support, and preparation of analysis-ready outputs. For pilot validation, a controlled masking experiment was performed on a representative tabular file from the dataset, in which 200 cells were manually removed and reconstructed by the GenAI pipeline. Comparison with the original reference values showed exact agreement for all evaluated entries, including both numeric and non-numeric fields. These findings indicate that GenAI can serve not only as an analytical aid but also as a practical tool for health data curation and controlled data completion in specialized biomechanical datasets. This result should be interpreted as a proof-of-concept evaluation under controlled masking conditions rather than evidence of general performance for real-world missing-data imputation.