I build machine learning systems that model and predict the real world.
Autonomous systems and multimodal detection in practice; world models and representation learning in research.
path consistency for latent world models
A self-supervised mechanism that penalizes disagreement between horizon-conditioned direct prediction and autoregressive rollout, aimed at reducing compound error in latent rollouts. Designed as an importable component rather than a standalone architecture.
acoustic early warning, and where it breaks
A Random Forest classifier on MFCC features reaching strong held-out metrics that then failed on real recordings while misflagging ambient noise. The case study is built around that gap: real-world test sets at fixed distances, cross-drone generalization, and an explicit account of what the training data does not contain.
a document research tool I actually use
A local RAG system for reading and querying research material: FastAPI backend, ChromaDB vector store, React frontend, running under Docker Compose. Built to make studying under load tractable.
I'm an ML engineer working on autonomous systems, multimodal learning, and prediction from sensor data. I like working across the full span of a problem — from data and experiments to evaluation and systems that actually run — rather than focusing on the model alone.
What draws me most is machine learning that tries to model the world, not just classify static data. My research interests are world models, representation learning, self-supervised learning, and long-horizon prediction; Anchor is where most of that thinking currently goes. In the long term, I want to work where research turns into systems that operate beyond benchmarks.
I study Data Engineering and Analytics at Rzeszów University of Technology and serve as Vice President of the Machine Learning Science Club, where my work spans research projects, events, and teaching.
Hertz New Technologies
Machine learning for autonomous systems and counter-UAS: multimodal detection, computer vision, prediction, and sensor data.
Umbris
Multimodal counter-UAS system fusing audio, RF, and vision, with a focus on edge inference and distributed sensing.
Machine Learning Science Club, Rzeszów University of Technology
Technical and organizational lead for research projects, events, and team development.
MTU Aero Engines Polska
Data warehouse modernization and aircraft engine analytics with dbt, Python, SQL, and Tableau.
Education Support Foundation, Aviation Valley Association
Taught electronics, programming, and STEM fundamentals in schools across Poland.
1st place
Built DetMesh, a multimodal counter-UAS system combining audio, RF, and vision for drone detection. The work later became Umbris.
1st place, Space category
AgloMeter, a data-driven analysis of transport exclusion and accessibility. Led the team as project manager.
1st place, Drones category
WildGuard, an early wildfire detection system using drone imagery and sensor data, with an alert dashboard for emergency services.