At Huuuge Games, I forecast KPIs with time-series models, optimize marketing calendars and events with heuristic search, and use explainable-AI methods to show what drives results. I also build LLM-based chatbots and AI-driven apps.
I co-authored mllm-shap, an open-source Python package for Shapley-value explainability of conversational and multimodal LLMs, presented at ACL 2026 System Demonstrations. It supports exact, Monte Carlo, and Leverage SHAP estimators, along with reproducible run archives.
At Data Fuze, I designed and optimized Snowflake data warehouses and integrated data sources using Python, Airflow, and dbt. I also built a real-time speech-to-text tool for doctor-patient interviews at MedWave and co-led the autonomous drone project for Droniada 2024.
