M.Sc. Business Analytics student at the Technical University of Denmark (DTU), Copenhagen.

I'm a data scientist who enjoys working with data in many forms: human biomechanics, conveyor-belt video feeds, maritime simulation, weather data. Across these domains I kept hitting the same gap: a good prediction doesn't tell you what to do. That's why I'm pursuing an M.Sc. in Business Analytics at DTU, where I focus on turning models into decisions that account for real-world constraints, costs, and uncertainty.

An end-to-end pipeline that turns raw movement into a graded technique assessment: framing a subjective judgement ("is this push-up good form?") as a supervised problem, engineering ~77 biomechanical features from pose data, and validating with subject-grouped cross-validation so the model generalizes to people it has never seen — not just unseen frames.

Problem framing Feature engineering Robust validation Python

A cross-lingual benchmark probing false-belief reasoning in modern LLMs across English, Polish, and German. Designed to expose where models are confidently wrong on second-order reasoning — the failure mode that matters most when systems are trusted with judgement.

Evaluation design LLM benchmarking Cross-lingual analysis

Real-time detection of fleeting arbitrage signals across equity and crypto markets. The hard part isn't the pipeline — it's reasoning about latency, windowing, and what counts as a genuine, actionable signal versus noise in a streaming setting.

Streaming analytics Signal detection Real-time systems

Integrating energy, climate, and economic data across all 27 EU countries into one analytical model, then surfacing the correlations that explain consumption and renewable-generation patterns. A study in making heterogeneous public data answer questions it was never structured to answer.

Data modelling Integration Analytical reporting

Optimization — convex optimization, gradient & quasi-Newton methods (L-BFGS), trust-region and subgradient methods, KKT/duality, LP/SQP, objective design with cost penalties.

Modelling & Statistics — regression and GLMs, time-series (AR/ARMA, regime-switching), MLE and Bayesian estimation, distribution-shift diagnostics, principled cross-validation.

Machine Learning — gradient boosting, SVMs, ensembles, clustering, reinforcement learning (PPO), and LLM evaluation & RAG.

Engineering — Python (numpy, pandas, scikit-learn, PyTorch, FastAPI), R, SQL, and Rust; streaming and batch data pipelines; cloud deployment on GCP.

Open to roles in data science, optimization, and operations analytics. Previously featured as an AI guest on TVP3 Warszawa. Always happy to talk through a hard problem — reach out.