NeuralNet project (PyTorch/MNE/Captum/Streamlit)
Built a deep learning model workflow for EEG-based motor imagery classification, including explainability outputs used to interpret model behavior. The experience involved preparing and analyzing multi-channel EEG inputs and visualizing model decisions rather than producing manual labels. It represents AI training and evaluation on scientific/biomedical signals. • Designed a hybrid CNN–LSTM–GRU PyTorch model for EEG motor imagery classification. • Processed and analyzed multi-channel EEG signals using MNE for feature extraction. • Applied Explainable AI via Captum to visualize model decisions. • Developed a Streamlit interface for real-time inference and model behavior demonstration.