01
Causal AI for drug discovery
Building treatment-effect and survival modelling pipelines from real-world data to move from population averages toward clinically meaningful, patient-level evidence.
Causal inferenceRWETarget discovery
✦
Pharmaceutical AI scientist · London
Evidence,
made individual.
I’m Katja (Ekaterina) Zilonova. I build AI that connects patient trajectories, causal inference and multimodal health data, turning complex evidence into better questions for drug discovery and precision medicine. I am sure we can make a difference by treating biology as data and harnessing today’s pace of technological progress.
Patient-level
evidence
evidence
Trajectories
Biology
Causal AI
01 / Selected work
Where computation meets
clinical consequence.
02
Learning patient trajectories
Using transformer-based foundation models to represent longitudinal EHR histories and reveal disease patterns across complex, multimodal cohorts.
TransformersEHRRepresentation learning
✦
03
Cardiovascular precision medicine
Leading AI and model evaluation in iCARE4CVD, connecting harmonised international heart-failure datasets with individualised prediction and treatment strategies.
Heart failureMultimodal AIIHI
04
Multimodal AI & biomarker discovery
Integrating EHR, medical imaging and proteomics for disease characterisation and biomarker discovery across international cohorts.
ProteomicsMedical imagingBiomarkers
✦
05
Medical imaging & regulated AI
Experience in ultrasound, CT and CBCT development and clinical validation under ISO 13485 quality management and ISO 14971 risk management, including evidence supporting MDR and FDA 510(k) regulatory submissions.
ISO 13485ISO 14971MDR & FDA 510(k)
✦
3data worldsElectronic health records and pharmacy claims · medical imaging · proteomics and biomarkers
30+international cohortsreal-world, trial & multimodal data
10+years across disciplinesmaths → medicine → AI
6people supervisedgraduate to postdoctoral level
02 / Journey
A career built
across boundaries.
From applied mathematics and ultrasound physics to regulated medical devices and causal AI in pharma, each chapter adds a different way of seeing the same problem: how to make health data genuinely useful.
London, United Kingdom
2024 to now
Senior Research Scientist
Novo Nordisk
AI, causal inference and multimodal real-world evidence for drug and target discovery; integration of EHR, imaging and proteomics; AI lead within iCARE4CVD.
2023 to 2024
Medical Data Scientist
NHS
Medical imaging data infrastructure for retrospective research cohorts, including pseudonymisation and medical-device software risk management under ISO 13485 and ISO 14971.
2021 to 2023
Image Quality Specialist
Philips Medical Systems R&D
AI-enabled CBCT image quality, radiation-dose reduction, and clinical and technical validation contributing to MDR and FDA regulatory submissions.
2019 to 2021
Postdoctoral Researcher
KU Leuven / UZ Leuven
Ultrasound reconstruction, beamforming and CT-derived device specifications for novel medical imaging systems.
03 / Talks & conference work
Ideas are shaped
in conversation.
2026European Society of Cardiology Congress, Munich
From guidelines to precision care: a personalised approach to heart failure management using heterogeneous treatment effectsCo-author
2026European Society of Cardiology Congress, MunichRobust cross center coronary artery segmentation on computed tomography angiography with limited local annotationsCo-author
2026University of Cambridge / Newton GatewayTowards precision medicine: causal inference opportunities and challenges in observational healthcare dataInvited talk · speaker
2026German Cardiology Congress / DGKStandardization and AI-readiness across 5 international heart failure datasets to support the development of AI modelsCo-author
2025Women in Data Science MaastrichtiCARE4CVD: A journey into harmonising and analysing multiple heart failure cohort studiesInvited talk
2024Single Voxel GroupXNAT PresentationSpeaker
2020IEEE International Ultrasonics SymposiumTransthoracic cardiac ultrasound imaging using a flexible transducer array: in silico feasibility studyPresenter
2020IEEE International Ultrasonics SymposiumA Novel 6 MHz Phased Array Piezoelectric Micromachined Ultrasound Transducer with 128 Elements for Medical ImagingCo-authored conference proceeding
20182nd International Conference on Mechanics (ICM)Cavitation dynamics in high-intensity focused ultrasound thermal therapyPresenter · Yilan, Taiwan
201813th SIAM East Asian Section Conference (EASIAM)Cavitation and corresponding temperature distribution in soft tissue during HIFU thermal therapyPresenter · University of Tokyo, Tokyo, Japan
2017International Conference on Biomedical Engineering (ICBMU)Bubble dynamics and corresponding heat deposition in soft tissue during high-intensity focal ultrasound thermal therapyPresenter · The Hong Kong Polytechnic University, Hong Kong
2015Fifth International Conference on Multiscale Modeling and MethodsSearch for strategy of treatment in the mathematical model of the interaction between microbial population and antibioticPresenter · Bauman University, Moscow, Russia
04 / Selected publications
Published across scales,
from bubbles to populations.
2026From guidelines to precision care: a personalised approach to heart failure management using heterogeneous treatment effectsEuropean Society of Cardiology Congress 2026 · congress abstract2026Robust cross center coronary artery segmentation on computed tomography angiography with limited local annotationsEuropean Society of Cardiology Congress 2026 · congress abstract2026Individualized Treatment Effect Estimation in Chronic Disease: Structural Challenges and a Path ForwardCausality for Decision Making Workshop · Conference on Uncertainty in Artificial Intelligence (UAI 2026) · OpenReview2026Standardization and AI-readiness across 5 international heart failure datasets to support the development of AI models: an iCARE4CVD use caseClinical Research in Cardiology · DGK conference abstract2021A Novel 6 MHz Phased Array Piezoelectric Micromachined Ultrasound Transducer (pMUT) with 128 Elements for Medical ImagingIEEE International Ultrasonics Symposium · conference proceeding2020Transthoracic cardiac ultrasound imaging using a flexible transducer array: in silico feasibility studyIEEE International Ultrasonics Symposium · conference proceeding2019Dynamics of bubble-bubble interactions experiencing viscoelastic dragPhysical Review E 99:0231092019Simulation of cavitation enhanced temperature elevation in a soft tissue during high-intensity focused ultrasound thermal therapyUltrasonics Sonochemistry 53:11-242018Bubble dynamics in viscoelastic soft tissue in high-intensity focal ultrasound thermal therapyUltrasonics Sonochemistry 40:900-9112018Managed Saffman-Taylor instability during overflush in hydraulic fracturingJournal of Petroleum Science and Engineering 162:513-5232016Optimal strategies in antibiotic treatment of microbial populationsApplicable Analysis 95(7):1534-1547
05 / Education
Three disciplines.
One through-line.
MSc Machine Learning & Data Science
Imperial College London
PhD Biomedical Engineering
National Taiwan University
MSc + BSc Applied Mathematics & Computer Science
Moscow State University