Computational Neuroscience Lakehouse (NeuroLab)

NeuroLab is our research data-engineering project: raw sleep EEG is organised into Bronze / Silver / Gold layers for a traceable, reproducible pipeline.

The data are public PhysioNet Sleep-EDF recordings. Raw EDF files arrive incrementally in Bronze via Auto Loader; Silver applies band-pass filtering, artefact detection and 30-second epoching and detects sleep spindles and slow oscillations; Gold aggregates time-domain, frequency-domain and topological features.

Topological features use methods such as persistent homology and complement classical time–frequency features.

To validate methods, we reproduced published sleep-oscillation research in Python: on a public intracranial EEG dataset we detect slow oscillations, spindles and ripples and analyse how they nest, using phase–amplitude coupling and cluster-based permutation tests.

The same methods power enterprise data platforms: layered modelling, data-quality checks, metadata management and reproducible computation.