AI/ML Engineer · PhD in Computer Science · Vienna, Austria
I work on GenAI backend systems and ML evaluation, with a research background in medical imaging. My engineering work covers event-driven inference, provider integrations and observability; my research focuses on patient-level evaluation, calibration and reproducibility.
- Production GenAI: built and operated a FastAPI/AWS backend for multimodal generation, integrated external AI providers, and implemented retries, tracing, cost monitoring and asynchronous inference workflows.
- Applied ML: developed and deployed churn and lifetime-value models, with model- and data-drift monitoring.
- Current work: building an internal knowledge assistant with TypeScript/Deno, PostgreSQL text retrieval, tool use and application-enforced source citations. This project is in progress.
- clinval-validator — clinical-ML evaluation tooling for patient leakage, split-sensitivity and subgroup degradation. Python computes the metrics; optional Claude reporting explains the findings. Demonstrated on synthetic/public-schema data.
- repro-survival-model-evaluation — agent-driven, CPU-only reproduction of an ICML 2026 paper on survival-model evaluation. On METABRIC at reduced scale: one claim reproduced, one graded partial in the repository.
- repro-instance-level-costs — agent-driven reproduction of an ICML 2026 paper on cost-sensitive classifier evaluation. The headline result reproduces on Jigsaw; cost-weighted training only partially; fine-tuned-model results were not reproduced.
- finance-bot — single-user Telegram finance bot built with Python, AWS Lambda and DynamoDB, with Terraform infrastructure, recurring bookings and CSV export.
- RFOCT — research/reference implementation of Random Forest of Optimal-Complexity Trees, the medical-image classification algorithm from my first-author 2023 paper in Cybernetics and Systems Analysis.
- ABPMHemodynamicCoupling — research pipeline for stress-linked blood-pressure analysis, with subject-level modelling, cohort statistics and a Streamlit review interface. Supports our IEEE ELNANO 2026 work.
PhD in Computer Science, Igor Sikorsky Kyiv Polytechnic Institute (2025). My doctoral work studied hierarchical ensemble classifiers for pathology diagnosis from medical images, with retrospective evaluation in liver fibrosis staging and cardiac imaging. I also worked on physiological-signal analysis for stress and cognitive workload in a collaboration with the University of Calgary (2022–2026).
ORCID · Google Scholar · Scopus
Python · TypeScript · SQL · FastAPI · PostgreSQL · AWS · Docker · scikit-learn · PyTorch