Vol. 01 — Portfolio · Washington, DCFiled under: Neuroscience · Signal Processing · Machine LearningOpen to research & engineering roles
Vol. 01 · Portfolio · Est. 2008

Computational Neuroscientist · ML Engineer

Turning complex neural data into clinical impact.

I am a Post-Doctoral Researcher at Georgetown University Medical Center, specializing in Computational Neuroscience, Signal Processing, and Machine Learning — from cardiac digital twins for first-of-their-kind clinical trials to decoding human speech with advanced neuroimaging.

16Years in research
2008 — 2024
05Research
institutions
Nature BMEFirst-of-its-kind
clinical trial
NIH T32TL1 — independent
fellowship

About

001 / 008

My work translates complex neural and physiological data into clinical impact — building patient-specific models, decoding fast neural dynamics, and engineering the pipelines that make both reproducible at scale.

Side note

Trained across neuroscience, biomedical engineering, and HPC, I sit at the seam between the wet lab and the cluster: designing the experiment, writing the firmware, and building the model.

I care about methods that survive contact with real, noisy, motion-corrupted data — and about cutting the distance between a recording and a decision.

Bio / Timeline

002 / 008
2024 -
Post-Doctoral Researcher
Georgetown University Medical Center
Computational models of speech processing.
2018 - 2024
PhD in Neuroscience
Georgetown University
Thesis on Neural Dynamics of Speech. NIH T32 & TL1 Fellow.
2019
Research Fellow
Lawrence Livermore National Laboratory
ML framework for TBI recovery trajectories.
2014 - 2018
Research Engineer
Johns Hopkins University
Cardiac 'Digital Twins' and clinical trials for arrhythmia.
2011 - 2013
MS in Biomedical Engineering
Virginia Commonwealth University
Thesis on cognitive-visuomotor multitasking.
2008 - 2011
BS in Biomedical Engineering
Virginia Commonwealth University
Senior Capstone: Real-time BCI system.

Research & Projects

003 / 008
01

Source-Space Toolbox for Speech EEG ↗

2024 -
Postdoctoral Researcher @ Georgetown University Medical Center

A stand-alone pipeline that extends the sensor-space PhD analyses into cortical source space: source estimation, multivariate decoding, multi-atlas parcellation and directed connectivity, built on MNE-Python and scikit-learn.

  • Source estimation: projects 64-channel EEG onto fsaverage cortex via a shared forward model (ico-5, 3-layer BEM) with dSPM minimum-norm and LCMV beamformer inverses, per-subject data-rank estimation, and low-memory streaming so whole cohorts run on one machine.
  • Decoding: sliding-window classification of ROI source activity with linear SVM, shrinkage LDA and elastic-net logistic regression; nested-CV tuning, vertex/PCA feature modes, pseudo-trial averaging, and spatial leakage correction (symmetric orthogonalisation, regression-based vertex correction).
  • Parcellation: 16 speech-network ROIs defined on three atlases (Desikan-Killiany, Schaefer-200, HCP-MMP 1.0) plus custom functional-localizer ROIs projected from volumetric NIfTI masks to the cortical surface.
  • Granger causality: parametric (Morf MVAR with Geweke spectral GC; pairwise, block and conditional), nonparametric (Wilson spectral factorisation) and state-space (DARE, MVGC-equivalent) estimators, validated against BSMART to machine precision, with time-reversed GC controls and model-order and window diagnostics.
  • Statistics and reproducibility: cluster permutation and TFCE for decoding, permutation task-vs-baseline tests with FDR for connectivity, two-stage subject-parallel and ROI-parallel runners, and theory chapters and technical reports documenting every methodological choice.
View Analysis Code ↗
NeuroscienceEEGSource LocalizationGranger CausalityMNE-Python
02

Neural Dynamics of Speech Perception & Production ↗

2018 - 2024
PhD Researcher @ Georgetown University

Investigation of computational models engaged during auditory and spoken word speech processing, with potential applications for neuroscience-based speech therapy. Built the full sensor-space analysis stack for a 64-channel EEG perception and overt-production study, from the behavioural paradigm through preprocessing, decoding, connectivity and statistics, in MATLAB (EEGLAB, FieldTrip, BSMART) and Python.

  • Developed methods for EEG image acquisition and decoding of very fast (millisecond range) neural dynamics involving significant motion artifacts.
  • Developed multivariate analysis pipelines using HPC clusters to analyze large neural image datasets (cut preprocessing time from 3 weeks to 2 days).
  • Led interdisciplinary team investigating evidence of internal inverse and forward models in the brain.
  • Secured independent funding (NIH T32 & TL1) and led grant proposals from concept to award.
  • Behavioural paradigm: perception and overt-production tasks in Psychtoolbox with hardware-synchronised EEG and audio triggers.
  • Preprocessing in EEGLAB with speech-artifact removal by blind source separation with canonical correlation analysis (BSS-CCA), so production epochs can be analysed through articulation.
  • Automatic speech-onset detection for production recordings: a centre-of-gravity acoustic detector with parameter search, validated against two human raters, giving accurate onset-locked epoch windows.
  • Decoding: sliding-window SVM classification (linear and RBF, grid-searched) and representational similarity analysis with cross-validated distances and multivariate noise normalisation, run whole-brain (searchlight) and per ROI.
  • Dynamic visualisation: static topographies and topographic movies on three anatomical views for decoding and RSA time courses.
  • Directed connectivity: moving-window parametric (MVAR) Granger causality with the BSMART toolbox, seed-to-target and pairwise, with band-resolved and baseline-normalised figures.
  • Non-parametric cluster-based permutation statistics (dependent-samples t, Monte Carlo) on first-level ERP data, with parameter sweeps to check cluster robustness.
NeuroscienceEEGHPCSignal ProcessingMATLABEEGLAB
03

Cardiac 'Digital Twins' & Arrhythmia Prediction ↗

2014 - 2018
Research Engineer @ Johns Hopkins University

Large-scale electrophysiology simulations to analyze and predict arrhythmia behavior using patient-specific models.

  • Built MR-derived, patient-specific cardiac models and ran large-scale EP simulations.
  • Delivered the first clinical trial using virtual heart EP simulations to inform ventricular tachycardia ablation planning (Nature Biomedical Engineering).
  • Led two industry-academic partnerships to productionize finite-element solvers for biologically realistic heart models.
Nature BME Paper ↗
CardiacModellingSimulationClinical Trial
04

Connectome-Based ML for TBI Recovery ↗

2019
Research Fellow @ Lawrence Livermore National Laboratory

Built an end-to-end machine learning framework to model recovery trajectories in traumatic brain injury patients.

  • Built connectome-based ML framework using 2-week post-injury fMRI from the TRACK-TBI cohort.
  • Processed functional connectivity matrices to extract graph/network features.
  • Trained elastic net, ridge, and gradient boosting models to predict Glasgow Outcome Scale-Extended (GOSE) at 3/6/12 months.
View Pipeline on GitHub ↗
Machine LearningfMRIPythonTBI
05

BCI & Real-Time Bio-Signal Processing ↗

2008 - 2013
Research Assistant / BS Capstone @ Virginia Commonwealth University

Design and implementation of non-invasive brain-computer interface technologies.

  • Designed visual stimulus-based BCI system integrated within a real-time EEG pipeline.
  • Wrote assembly-level firmware for deterministic timing and achieved four-class control for on-screen navigation.
  • Developed algorithms for motor imagery BCI, validating timing determinism and signal integrity.
  • Conducted empirical evaluations to optimize accuracy, false-activation rate, and user training time.
BCIEEGEmbedded SystemsReal-time

Featured Talks

004 / 008
2024

Neural Dynamics of Speech Processing

Georgetown University Neuroscience Seminar
Washington, DC

Overview of my doctoral research on how the brain processes spoken language using rapid neural dynamics.

2019

Machine Learning for TBI Recovery

LLNL Summer Student Symposium
Livermore, CA

Presenting a framework for predicting recovery trajectories in TBI patients.

Blog

005 / 008

To do research is to teach myself and to teach the world whatever I discovered, or whatever I and my colleagues discovered together.

— Steven Strogatz, 3Blue1Brown podcast, 2026

Publications

006 / 008

Hobbies / Personal Projects

007 / 008

Side projects — usually at the intersection of code, data, and curiosity.

H01Source estimation sandboxFull-stack toolbox for EEG/MEG source localization, decoding, and functional connectivity.EEG
H02Android App: Training LogAndroid workout and diet tracker where how you feel each morning sets how hard the day's session is. Diet plan pre-made and imported. Macro tracker + weekly analytics. Written in TypeScript; all data stays on the phone.Android

Contact

008 / 008

Let's build something impactful.

nikolov[dot]phd[at]gmail[dot]com

Coffee

Vol. 02

EspressOnly.

↻
000° / 720°Closed
Pressure · barFlow · ml/s0.0 s