Brain–Computer Interface Signal Pipelines

Overview

Work in Columbia’s Brain–Computer Interfaces Lab (Fall 2025): collecting EEG data and building classification pipelines for multiple paradigms used in non-invasive BCIs.

Technical work

• Motor imagery: Common Spatial Pattern (CSP) filtering and classifier tuning — 87.81% accuracy on our held-out evaluation
• P300: Event-related potential detection pipeline for spellers / selection tasks
• SSVEP: Steady-state visual evoked potential analysis for frequency-coded targets

Why it matters

BCI sits at the intersection of signal processing, neuroscience, and human-computer interaction — the same “signals → model → action” loop I use in genomics and health analytics, applied to neural data.

Stack

Python, MNE-style preprocessing patterns, scikit-learn, feature extraction and cross-validation for small-N EEG datasets.