Abdullah S.

Struggling with noisy EEG, unreliable classifiers, or brain data you can't make sense of? I build EEG and fMRI signal processing and machine learning pipelines that turn raw brain signals into results you can trust and defend. I'm a neurotechnology ML engineer with hands-on experience across the full EEG workflow, from recording and preprocessing to deep learning models and analysis tools. My research focuses on multimodal deep learning for sleep staging using simultaneous EEG-fMRI data. CLIENTS TYPICALLY HIRE ME TO - Clean and preprocess EEG recordings: filtering, ICA artifact removal, signal quality checks - Build ML and deep learning models for cognitive state, sleep stage, or BCI classification - Analyze ERPs, band power, and spectral features for research studies - Set up research-ready data pipelines: BIDS formatting, event markers, synchronized task recording - Diagnose why a model or dataset isn't producing reliable results SELECTED WORK - Sleep staging with simultaneous EEG-fMRI: multimodal DL model using 32-channel EEG recorded at 5000 Hz alongside fMRI. - Cognitive state classification: EEGNet models for multi-class classification, supported by a cognitive task battery (Flanker, N-Back, Stroop, Math, PVT) with millisecond-accurate synchronized event markers. - Wearable Neurotech: Designed a wearable neurotech devices and its complete signal pipeline, including noise correction for weak signals and real-time electrode quality monitoring - Scalp EEG validation: validated EEG acquisition on a 32-channel EEG glasses system with OpenBCI Cyton at T7, T8, F7, and F8 - ERP analysis: Auditory oddball paradigm with N1-P2 extraction and jitter correction for accurate event timing - Analysis tools: desktop applications with real-time spectral visualization and marker synchronization checks. TOOLS Python: MNE-Python, PyTorch, scikit-learn, SciPy, BrainFlow MATLAB: EEGLAB Datasets and formats: EESM19, EESM23, EDF, BDF, .set, BIDS Devices: OpenBCI, ADS1299-based systems, BLE EEG streaming HOW I WORK You get clean, documented, reproducible code, a clear explanation of every processing decision, and regular progress updates. Because I've worked with EEG at every level, from electrode contact to classifier output, I can tell whether a problem comes from the model, the preprocessing, or the recording itself. Send me a short description of your data and your goal, and I'll tell you how I would approach it.