EEG/BCI Data Scientist · Computational Neuroscience · Scientific ML
I am a Ph.D. Candidate at UMass Boston working with Prof. David Degras. I study a practical problem: EEG/BCI models often work in one session and fail in the next. My work turns that problem into measurable pieces: stable signal representations, cross-session adaptation, drift diagnostics, and software that makes the experiments easier to rerun and inspect.
Featured Software & Systems
These projects follow one chain: build stable EEG features, adapt across sessions, compare source-session strategies, diagnose drift, and stress-test the geometry with simulation. The goal is not to hide messy EEG behavior, but to make it visible enough to support a better modeling decision.
Streamlit dashboard for MI-EEG drift diagnostics across multiple shift metrics, DA methods, feature families, and fixed-reference monitoring protocols.
Stack: Python, Streamlit, Plotly, pandas, statsmodels, scikit-learn, DA4BCI-Python, MOABB
Python package and CLI for running cross-session EEG domain-adaptation experiments across multiple source-session utilization strategies.
Stack: Python, MOABB, MNE, pyriemann, scikit-learn, DA4BCI-Python, cross-session EEG benchmarks
R toolkit for train/test-safe EEG feature extraction across spatial-filter, covariance, spectral, dynamical, and avalanche feature families.
Stack: R, EEG feature extraction, CSP/FBCSP/FBCSSP, Riemannian geometry, MVAR/MSVAR, BCI decoding
Python backend for EEG/BCI domain adaptation, shift metrics, SPD geometry, preprocessing, and drift detection.
Stack: Python, NumPy, SciPy, scikit-learn, POT, SPD geometry, EEG/BCI domain adaptation
R package for synthetic EEG tensor simulation, SPD covariance augmentation, fidelity auditing, and manifest-based replay.
Stack: R, EEG simulation, SPD covariance geometry, Riemannian augmentation, fidelity audit, reproducibility
Interactive benchmark dashboard for comparing source-session utilization strategies across MAP, DWP, MMP, and BDP pipelines.
Stack: Python, Streamlit, Plotly, pandas, NumPy, cross-session EEG benchmarks, multi-source domain adaptation
Research
I study why EEG pipelines fail across sessions and how to diagnose, benchmark, and adapt them with measurable criteria.
Methods
My work links distribution shift to practical choices about pooling, source selection, adaptation, retraining, and feature recalibration.
Engineering
I build R and Python tools for feature extraction, whitening, domain adaptation, benchmark auditing, drift dashboards, and simulation.
Research Program
A matched comparison of global and selective source-session pooling strategies for cross-session EEG transfer.
A geometric framework for separating raw sensor drift from feature-space distortions in MI-EEG decoding.
A linear-first decision rule for balancing MI-EEG decoding accuracy, stability, and computational cost.
Writing
A working note on using regime-switching linear state-space models for non-invasive brain-signal decoding.
A direct guide to reading when EEG domain adaptation is useful and when it causes negative transfer.
A practical note on Wasserstein, MMD, and Energy Distance for quantifying session-to-session shift in EEG pipelines.
A practical view of how multiple source sessions can help or hurt cross-session EEG transfer.
Project Archive
Static browser app for teaching and inspecting domain-adaptation geometry with 2-D toy data, method animations, metrics, a...
Desktop app for sample-paced BCI replay with four motor-imagery pipelines, Euclidean Alignment, and live inference visuali...
Python package mirroring TensorEEG's simulation, augmentation, SPD geometry, fidelity audit, and manifest replay APIs.
Rank: 98/2767 (Top 4%) | Silver Medal Developed a deep learning pipeline using EfficientNet and Weighted Ensembling to cla...
Contact
I am open to computational neuroscientist, research scientist, data scientist, and scientific ML roles where rigorous evaluation, EEG/BCI systems, and production-quality research tooling matter.