Kyriakos Schwarz

Cheminformatics and machine learning for reaction intelligence and compound design.

I collaborate with medicinal chemistry, CADD, and Computational Sciences teams on cheminformatics and AI for drug discovery. My work includes machine learning for atom mapping and reaction class annotation, analysis of a new reaction database at Roche, an ML-guided compound selection tool with a user-friendly GUI, and a golden set of building blocks for accelerated drug discovery.

Production machine learning and data interoperability for customer analytics.

I worked across the full Big Data pipeline, from feature engineering in Oracle SQL to deploying and maintaining production AI models through MLOps processes. I also helped transition the team from a traditional Data Warehouse to a multi-cloud Databricks solution on Microsoft Azure.

Medical AI research in computational pharmacology and data interoperability.

I developed an explainable PyTorch Transformer model to predict medication side effects from diverse data, including chemical structures. I analyzed UK Biobank data to study drug-disease relationships and built a Graph Neural Network with PyTorch and RDKit for drug synergy prediction.

Full-stack web development and CMS engineering in an Agile/Scrum environment.

I developed front-end applications with HTML, CSS, and JavaScript, configured Magnolia CMS, extended its functionality in Java, and built REST and SOAP services. I used Git, Maven, Tomcat, Jenkins, and Docker while applying CI/CD best practices.