I lead the design of genomic evidence systems that power drug discovery and clinical decisions with high-confidence human genetic insights.
Profile
I am a statistical geneticist and genomic platform lead operating at the intersection of genomics, AI, and translational medicine. At Pheiron, I lead the design of biobank-scale genomic evidence systems that translate human genetic data into actionable inputs for drug development and clinical decision-making.
My work combines statistical genetics, causal inference, and multi-omic integration, with an emphasis on building robust, production-grade platforms. I have led efforts spanning cross-genomic regulation—most notably tissue-specific mitochondrial–nuclear gene regulation—and large-scale graph-based integration of biomedical knowledge. I am drawn to problems where scientific rigor, system architecture, and translational impact are addressed within a unified integrated framework.
These systems are designed to support causal interpretation, target validation, biomarker development, and clinical prioritization at decision-making scale.
I work fluently across R (Bioconductor, Shiny), Python (Polars, Limix), and C/C++ (PLINK, REGENIE, LDAK, GCTA), and have led projects using Neo4j, RDF/SPARQL, and cloud‑ready analytics platforms.
On the platform side, I build queryable data catalogs on Apache Iceberg/Parquet with SQL/DuckDB analytic layers and REST interfaces for gene, variant, and phenotype‑level lookups, backed by production‑grade QC, validation, and monitoring. My experience spans DNAnexus, AWS/S3, Docker, and reproducible workflows for biobank‑scale analyses, together with methods expertise in GWAS, PheWAS, eQTL/ct‑eQTL mapping, PRS, rare‑variant burden tests, fine‑mapping, and causal follow‑up (colocalization, mediation, Mendelian randomization).
Beyond tools, I focus on reproducibility, robustness, and aligning computational strategy with scientific and translational objectives.
At Pheiron, I lead the design of genomic evidence systems that support clinical translation and drug discovery, from data ingestion to decision‑facing outputs. My work spans integrating large biobank and disease cohorts (including UK Biobank, FinnGen, and GTEx), building scalable GWAS/eQTL and burden‑testing pipelines, and shaping platform architecture that combines AI and statistical genetics to produce causal, decision‑grade readouts from population‑scale multi‑omic data.
Joined the Computational Neurobiology Group at Helmholtz Computational Health Center to modernize the Alzheimer’s Disease Atlas by integrating large‑scale omics data into a Neo4j‑backed knowledge graph. I designed unified scoring schemes and harmonization workflows for multi‑omic and phenotypic data, enabling network‑ and AI‑driven prioritization of therapeutic targets and candidate drugs.
Funded through the highly competitive Munich School of Data Science
(MUDS), I worked as a scientific researcher at the
Institute of Translational Genetics
, where I led work on mitochondrial–nuclear regulation across 40+ human tissues, building multi‑tissue atlases of gene expression and mito‑nuclear control.
This work produced first‑author and co‑first‑author manuscripts on mito‑nuclear gene expression regulation and tissue‑specific mtDNA heteroplasmy (both under revision at Nature Communications), and involved designing large‑scale WGS/RNA‑seq pipelines, causal inference workflows, and cross‑disciplinary collaborations, alongside co‑organizing two MUDS retreats and their scientific programs.
I was awarded a fellowship from the Data Science for Social Good Foundation (DSSG), during which I collaborated with Munich's Machine Learning Center (MCML), the statistical department of LMU, and Bavaria's state Ministry of the Interior (BMI).
Our collaboration aimed to develop a web application for identifying optimal water extraction points for firefighting purposes. The project involved integrating statistical mapping, machine learning techniques, and interactive maps.
Additionally, we implemented routing algorithms, containerized the application using Docker, deployed it on AWS, and presented our findings to Germany's federal parliament.
Instructor, Human Genetics of Complex Traits
October 2022 - March 2023
Technical University of Munich
I developed and configured educational material for TUM's Medicine department's Human Genetics of Complex Traits course (ME 1660-P).
In addition, as part of our team, I served as a tutor in lectures and workshops, where we covered a diverse range of topics including UNIX, quality control, association testing, meta-analysis, and polygenic risk scores.
These roles, helped me facilitate a comprehensive learning experience for students, which later inspired me to pursue a professional certificate in Teaching in Higher Education from TUM ProLehre Media & Didactics.
Genomic Data Science Intern
October 2019 - April 2020
Helmholtz Munich
Implemented my master thesis at the Institute of Translational Genomics (ITG), where I developed a quality control pipeline to analyze pre-processed data. The pipeline ensured accuracy during merging, corrected for population structure, imputed missing data, and identified biases.
Focused on improving the quality of genome-wide association studies (GWAS) data, I processed peripheral whole blood samples collected from 250 patients across four cohorts undergoing total joint replacement surgery. The project involved merging cohorts from different array versions, necessitating specialized handling of genetic data assumptions.
Education
Dr. rer. nat.
June 2020 - December 2024
Technical University of Munich, DE
During my doctoral program in experimental medicine and health sciences, I developed a computational pipeline to analyze mtDNA-nucDNA interactions across 48 tissues, identifying discrepancies related to psychiatric conditions. Leveraging machine learning approaches, we revealed regulatory relationships between mitochondria and the nucleus, pinpointing causal genomic variations, particularly in the central nervous system tissues. This research sheds light on tissue-specific mito-nuclear interactions, crucial for understanding health and disease.
M.Sc.
October 2018 - April 2020
Thessaly University, GR
Joined the department of computer science and biomedical informatics, where I took focused courses on applied machine learning, ontology engineering, biosensors and digital imaging. My master's thesis centered around identifying biased individuals and SNPs, statistical imputation and concatenating cohorts from different Illumina genotyping arrays.
B.Sc. (Hons)
October 2012 - April 2018
Piraeus University, GR
During my 4-year degree in statistics and actuarial mathematics, I gained a solid educational background in probability theory, linear algebra, bayesian statistics, biostatistics, risk management and stochastic processes.
In my bachelor thesis, I explored the application of fitting probability distributions to empirical data, where I analyzed various techniques to optimize model selection and enhance the accuracy of predictive systems.
By using linear mixed models (LMMs) to conduct expression quantitative trait locus (eQTL) mapping, we explored the crosstalk between mitochondria and the cell nucleus across 48 different tissues. Through the integration of our identified eQTLs with various molecular data, we applied a series of analyses, including statistical colocalization, causal inference, graph neural networks, and over-representation analysis. These methods helped us identify disease genes, variants, and ontological pathways. Our study shows potential for unveiling new therapeutic approaches aimed at a wide range of diseases, from metabolic disorders to neurodegenerative conditions.
During the conference, I showcased my poster on the interaction between mitochondrial and nuclear communication in regulating tissue-specific gene expression. In addition to receiving excellent feedback, I had the privilege of meeting numerous outstanding scientists and fostering valuable discussions.
This project aims to develop a web application to help firefighters determine the areas covered with existing and planned hydrants. The application identifies accessible zones utilizing hydrant location or natural water sources, pinpoints nearby water sources for firefighting purposes, and calculates elevation disparities between the fire location and surrounding water sources.
Data Science for Social Good, DSSGxMunich, MCML, Machine Learning, BERD, BYTE, Bavarian Ministry for Digital Affairs
I have been involved in the curation of a workshop at TUM that covers key subjects in human genetics and genomics. The course included various topics such as tools for genome-wide association studies (GWAS), UNIX command line, association testing, meta-analysis methods, rare genetic variations, molecular QTL mapping, bioinformatics databases, and polygenic scores. Moreover, I led the creation of interactive Jupyter lab exercises during the workshop's hands-on sessions, allowing participants to apply their knowledge of the course content to real-world data.
In this study, we comprehensively characterize the landscape of common mtDNA heteroplasmy and mtRNA modifications across 49 human tissues assayed with bulk RNAseq in GTEx v8. As mtDNA heteroplasmy and mtRNA modifications are tissue-specific, we aim to establish a robust framework for identifying and testing both types of variations from tissue-specific RNAseq data. our study provides the first comprehensive and tissue-specific description of the apparent mtDNA heteroplasmy, their relationship with donor age, and their regulatory effects on mitochondrial gene expression.
During the conference, I showcased my poster on the interaction between mitochondrial and nuclear communication in regulating tissue-specific gene expression.
We presented research findings related to mitochondrial DNA (mtDNA) heteroplasmy and its association with ageing and gene expression patterns, shedding light on the potential functional consequences of mtDNA variation within different tissues.
This thesis explores the crucial importance of quality control (QC) procedures in genome-wide association studies (GWAS), which are vital for ensuring the trustworthiness of genetic associations. It focuses on integrating cohorts from different array versions, tackling the computational hurdles and genetic assumptions present in Illumina genotyping data. By employing thorough QC filtering and imputation methods, the thesis introduces a strong framework for detecting outlier individuals and SNPs, ultimately generating a top-notch dataset suitable for precise association testing.
STARS GAME is a project that is transforming the way science education is approached for 10-13-year-olds. As part of the game's design team, we secured and executed an EU commission-funded initiative. Through design thinking, we developed an educational escape room, integrating science into school curricula across 30 institutions in four European countries. Our main objective was to connect theoretical knowledge with practical application, utilizing game-based learning to nurture curiosity and improve critical thinking abilities.
Science education, Design Thinking, Digital Escape Room, Game-based learning, Erasmus, EU
