← Back to ML
ML /
ML Engineer @ NeuroTech @ Berkeley · Jan 2026 — Present

NeuroTech EEG Gait Decoder

TLDR

  • Built a non-invasive EEG gait-decoding pipeline for BCI-driven mobility systems, moving from angle-only regression toward multi-task gait understanding.

  • Preprocessed noisy EEG with artifact removal, bandpass filtering, per-channel normalization, and overlapping 1-second windows aligned for gait prediction.

  • Trained a TCN over 64-channel windows to predict heel_logit, phase_logit, and angle_pred, using cyclic phase features and staged training to strengthen ankle-angle prediction.