Data Science MSc student at FER, University of Zagreb
Working toward a career as a Research Engineer / Applied Scientist
focused on NLP, foundation and multimodal models, information retrieval,
and agentic AI.
I am interested in combining rigorous ML research with production-grade engineering to build systems that use external knowledge and tools while remaining reliable, interpretable, safe, and useful.
-0078D4?style=for-the-badge&logo=microsoftoutlook&logoColor=white){kind=icon}
|
Spaniverse In development |
Benchmark and tooling for evaluating LLMs on span-level NLP tasks, supported by reproducible Wikipedia and Wikidata knowledge-base workflows. |
| CareFree ↗ | Student psychological-support platform exploring responsible, privacy-aware, and human-in-the-loop use of LLMs in a sensitive domain. |
| Lumen Recommender ↗ | End-to-end B2B recommendation system with hybrid models, temporal evaluation, interpretable segmentation, and a FastAPI serving layer. |
| SemEval 2025 ↗ | Multilingual role classification using XLM-RoBERTa, span representations, and hierarchical prediction constraints. |
- BSc thesis: developed an end-to-end entity-linking system for Croatian, including NER ensembling and candidate-retrieval analysis on the CRONEL dataset.
- Worked on multilingual and span-level NLP using transformer models, structured representations, and hierarchical prediction methods.
- Experience designing reproducible ML experiments with meaningful metrics, strong baselines, careful error analysis, and interpretable results.
- Built end-to-end ML and software systems spanning data pipelines, model development, APIs, testing, containers, and deployment.
Also used when needed: Pandas, NumPy, XGBoost, SHAP, Apache Spark, BigQuery, Google Cloud, Django, TypeScript, Next.js, and D3.js.
