0 0 1
(CA)
INDEX

Dear colleagues, collaborators, and curious minds — welcome to a professional presence built at the intersection of clinical neuroscience and machine learning. Not a brochure. A living record: of EEG signals decoded, algorithms evaluated, and knowledge contributed at the frontier of pediatric neurocritical care.

Dynamic Data Scientist with deep domain expertise as a Clinical Neurophysiologist at the University Children's Hospital Zürich (KISPI). Passionate about Healthcare Foundation Models · XAI · Generalizability · Clinical Biomarkers · Precision Medicine.

§ What you will find here
001
Projects & Case Studies
From M.Sc. thesis ML pipelines to EEG annotation infrastructure and real-time neuromonitoring systems. Method, stack, outcome, and status documented.
→ View Projects
002
Research & Publications
Peer-reviewed publications and conference abstracts in EEG analysis, burst suppression detection, and unsupervised deep learning for clinical neurophysiology.
→ View Archive
003
Curriculum Vitæ
Clinical, research, and technical experience across KISPI, USZ, and UT Dallas. Education, certifications, and language proficiency — typeset as a living document.
→ View CV
004
Contact & Colophon
Open to research collaborations, clinical informatics projects, and interdisciplinary inquiries at the intersection of neuroscience and data science.
→ Get in Touch
is analyzed — signal by signal, patient by patient
is generalized — validated across heterogeneous and unseen data
is documented — peer-reviewed, open, attributed
not arriving — is not an option
At the time of launch,
with full archival rigor —
Carlos Arzaga
Data Scientist · KISPI · Zürich
0 0 3
Research Archive
Publications · Conference Abstracts · Posters
INDEX
2025

An Unsupervised Deep Learning Algorithm for Burst Suppression Detection in Pediatric EEG Data: Assessing Generalizability

Arzaga, C., Staubli, O., Birbaumer, M., Ramantani, G., & Keller, E.
Clinical and Translational Neuroscience · Vol. 9, No. 4, Article 45 [Abstract P136]
DOI: 10.3390/ctn9040045 · 6th Swiss Congress of Clinical Neurophysiology, 2025

Assessed the generalizability of a pre-validated unsupervised algorithm (BSUPP) for Burst Suppression pattern (BSP) detection on pediatric continuous EEG data (~12 hours, 4 PICU patients, University Children's Hospital Zürich). Developed a novel ensemble architecture combining BSUPP with a denoising Variational Autoencoder (dVAE), improving AUROC from 0.41 to 0.53 — a 30% increase — with perfect recall of 1.00 for burst detection on single-patient inference.

Journal Abstract Peer-Reviewed EEG Deep Learning Pediatric Neurocritical Care
2025

An Unsupervised Deep Learning Algorithm for Burst Suppression Detection in Pediatric EEG Data: Assessing Generalizability [Poster P136]

Arzaga, C., Staubli, O., Birbaumer, M., Ramantani, G., & Keller, E.
6th Swiss Congress of Clinical Neurophysiology
Affiliations: University Children's Hospital Zürich (KISPI) · HSLU · University Hospital Zürich (USZ)

Poster presentation evaluating a dVAE+BSUPP ensemble architecture for automated BSP detection. Phase I tested the BSUPP surrogate model on ~12 hours of unseen pediatric cEEG data (4 PICU patients, status epilepticus), revealing limited direct generalizability (mean AUROC = 0.41). Phase II developed the novel ensemble through extensive hyperparameter experiments, achieving AUROC = 0.53 — a 30% improvement — with perfect burst recall of 1.00 on single-patient inference. Highlights unsupervised deep learning's potential for real-time automated BSP detection in clinical contexts.

Conference Poster Presented dVAE Burst Suppression Generalizability
2025

An Unsupervised Deep Learning Algorithm for Burst Suppression Detection in Pediatric EEG Data: Assessing Generalizability

Arzaga, C.
M.Sc. Thesis — Applied Information and Data Science
HSLU – Lucerne University of Applied Sciences and Arts · Feb 2023 – Sep 2025

Master's thesis investigating the generalizability of unsupervised ML algorithms for burst suppression pattern detection in pediatric cEEG data. Conducted in collaboration with the ICU Cockpit Research Group at USZ. Built a Label Studio annotation pipeline, curated and pre-processed EEG signals from 4 PICU patients with status epilepticus, and developed the dVAE+BSUPP ensemble architecture through hyperparameter optimisation experiments. Addresses the critical gap that fewer than 1% of 10,462 healthcare ML algorithms reported external validation.

M.Sc. Thesis HSLU Unsupervised ML EEG Signal Processing
§ Research Interests

Healthcare Foundation Models  ·  Explainable AI (XAI)  ·  Generalizability of Clinical ML Models  ·  EEG Biomarkers  ·  Precision Medicine  ·  Neurocritical Care Informatics  ·  Pediatric Epilepsy

0 0 2
Projects
Applied Work · Tools · Case Studies
INDEX

Applied research outputs, clinical data infrastructure, and ML systems. Each project is a documented artefact — method, stack, outcome, and status recorded with archival precision.

Completed
001

dVAE+BSUPP Ensemble — M.Sc. Thesis

Novel unsupervised deep learning ensemble combining a denoising Variational Autoencoder with the BSUPP algorithm for burst suppression detection in pediatric cEEG. Achieved 30% AUROC improvement with perfect burst recall. KISPI · HSLU · USZ.

PyTorch Python EEG dVAE W&B
Ongoing
002

EEG Annotation Pipeline — ICU Cockpit

Built a new end-to-end EEG data annotation pipeline using Label Studio for the ICU Cockpit Research Group at USZ. Curated and pre-processed continuous EEG signals from neurocritical patients; evaluated performance of multiple burst-suppression detection models.

Label Studio Python scikit-learn Docker
Archived
003

EEG + fNIRS Neurocritical Monitoring

Executed EEG and fNIRS studies on neurocritical patients at USZ. Supported PI research on neurocritical biomarkers, managed all resulting data, and developed study design for integrating commercial EEG/NIRS systems via systematic literature review.

EEG fNIRS REDCap Python
Archived
004

Intraoperative Neurophysiology Monitoring

Analysed real-time EEG and single-cell data for DBS implantation (Parkinson's/Epilepsy) at USZ. Previously executed comprehensive IOM strategies in the US — SSEPs, AEPs, VEPs — and developed training protocols for new neurophysiology staff at Monitoring Concepts.

EEG SSEPs DBS IOM
Ongoing
005

Research Identity System — This Portfolio

A brutalist-archival web portfolio for documenting scientific identity, publications, and projects. OULIPO constraint-driven design applied to professional presence. Built in the browser, deployed via GitHub + Vercel.

HTML CSS JS GitHub Vercel
Archived
006

Memory Biomarkers — Aging & Memory Lab

Research assistant at UT Dallas Aging and Memory Research Laboratory. Executed electrophysiology and genetics protocols to identify memory biomarkers in rodent tissue. Managed tissue histology, blood collection, and data for doctoral student thesis.

Electrophysiology Histology Genetics
0 0 4
Curriculum Vitæ
Living Document · Updated March 2026
INDEX

Carlos Arzaga

Data Scientist · Researcher · Clinical Neurophysiologist

Dynamic Data Scientist and researcher with deep domain expertise as a Clinical Neurophysiologist. Adept at utilising machine learning algorithms to analyse complex biomedical data, driving actionable insights, and impactful research. Passionate about advancing healthcare technologies through innovative solutions and interdisciplinary collaboration.

carlos.arzaga@kispi.uzh.ch +41 76 805 25 21 github.com/carzaga1 DOI · Publication
§ Education
Feb 2023 — Sep 2025
M.Sc. Applied Information and Data Science
HSLU – Lucerne University of Applied Sciences and Arts · Lucerne, Switzerland
Thesis: "An Unsupervised Deep Learning Algorithm for Burst Suppression Detection in Pediatric EEG Data: Assessing Generalizability." Conducted in collaboration with the ICU Cockpit Research Group, University Hospital Zürich.
Aug 2016 — May 2018
B.Sc. Neuroscience
University of Texas at Dallas · Dallas, TX, USA
Undergraduate research in electrophysiology and memory biomarkers at the Aging and Memory Research Laboratory.
§ Experience
Feb 2024 — Present
Neurophysiology Specialist — Neurology | EEG/Epilepsy
University Children's Hospital Zürich (KISPI) · Zürich, Switzerland
Performed, analysed, and assisted in diagnosis of outpatient, inpatient, and long-term (Telemetry) EEGs. Performed evoked potentials (SSEPs, AEPs, VEPs). Served as clinical annotator for EEGs supporting ongoing epilepsy research studies. Assisted as educating preceptor to new rotating medical residents.
Aug 2023 — Jan 2024
Visiting Graduate Researcher — ICU Cockpit Research Group
University Hospital Zürich (USZ) · Zürich, Switzerland
Project: Evaluation of Burst-Suppression Detection Models. Evaluated ML models for automatic burst-suppression detection in EEG signals. Built a new data annotation pipeline using Label Studio. Curated and pre-processed EEG signals; evaluated performance of several detection models.
Nov 2020 — Jan 2024
Neurophysiology Specialist — Neurology | EEG/Epilepsy
University Hospital Zürich (USZ) · Zürich, Switzerland
Analysed real-time EEG/single-cell data for DBS implantation (Parkinson's/Epilepsy). Supported PI research on neurocritical biomarkers using EEG and fNIRS systems. Executed EEG + fNIRS studies on neurocritical patients. Developed study design for integrating commercial EEG/NIRS systems via literature review.
Aug 2018 — Sep 2020
Surgical Neurophysiologist
Monitoring Concepts (US IOM) · Dallas, TX, USA
Analysed neurodiagnostic tests to maintain nervous system integrity during surgery. Executed comprehensive intraoperative monitoring strategies with surgical/anaesthesia teams. Developed training protocols and served as educational preceptor to new staff.
Aug 2017 — Dec 2018
Research Assistant — Aging and Memory Research Laboratory
University of Texas at Dallas · Dallas, TX, USA
Executed electrophysiology/genetics to find memory biomarkers in rodent tissue. Managed tissue histology, blood collection, and data for doctoral student thesis. Managed housing and care for an aging rat colony, ensuring research protocols.
Apr 2015 — Aug 2018
Anesthesia Technician — Anesthesiology | Surgery
Baylor Scott & White Medical Center · Plano, TX, USA
Provided critical support to anesthesiologists for surgical procedures and emergencies, including invasive monitor insertion (CVC/arterial) and difficult airway management. Prepared, calibrated, and troubleshot anesthesia delivery systems. Operated Cell Saver equipment.
§ Technical Skills
Programming & DevOps
Python · R · HTML
Git · Bash · GitLab
GitHub · Docker · MLOps
AI & Machine Learning
PyTorch · TensorFlow
Keras · scikit-learn
Weights & Biases
Time Series Analysis
Databases & BI
MySQL · MongoDB
REDCap
Tableau · Power BI
§ Certifications
Active
Deep Tech Talent Certificate (EIT)  ·  Board Certified in Intraoperative Neurophysiological Monitoring (ABRET)  ·  Good Clinical Practice Modules 1 & 2 (USZ)  ·  Basic Life Support (KISPI)
§ Languages
English (Native / bilingual)  ·  Spanish (Native / bilingual)  ·  German (Professional working proficiency)
§ Personal Interests
Traveling  ·  Analog Film Photography  ·  Baking  ·  Run Clubs  ·  Basketball  ·  Gravel Cycling
↓ Export as PDF
0 0 5
Contact
Reply Slip · Colophon · Method
INDEX

Send a message

§ Direct Contact

Email
carlos.arzaga@kispi.uzh.ch
Phone
+41 76 805 25 21
Location
Zürich, Switzerland

§ Connect
§ Colophon

Typeset in Libre Baskerville (display), Courier Prime (body), Source Code Pro (labels), and Homemade Apple (voice). Designed constraint-first — in the OULIPO tradition. Built in the browser. Deployed via GitHub · Pages.

§ Philosophy

"The constraint is not a limitation — it is the structure that makes the work legible." Archival clarity over decorative noise.

At the time of launch — with full archival rigor — Carlos Arzaga