Curriculum vitae · Timur Kulenović

Work Experience

SportradarLjubljana, Slovenia

Machine Learning Engineer2023 — Now

Synthesizing real-time market data and betting liabilities to automate optimal odds compilation.
Implementing player profiling models enabling automated account limits and fraud detection.
Engineering automated trading triggers to monitor liabilities and manage house exposure.

Telekom SlovenijeLjubljana, Slovenia

Data Scientist2020 — 2022

Developed ML models predicting the probability of customers leaving the company — churn models.
Clustering of IPTV users for the purpose of segmentation.

Faculty of Computer Science and Informatics, University of LjubljanaLjubljana, Slovenia

M.Sc. in Data Science and Computer ScienceOct. 2020 — Ongoing

Data Science master program is part of DataScience@UL-FRI and focuses on data analysis, statistics and machine learning.

Faculty of Computer Science and Informatics, University of LjubljanaLjubljana, Slovenia

B.Sc. in Computer ScienceOct. 2017 — Sep. 2020

Analysis of home advantage in sportsMaster Thesis · 2023

Addressed home advantage in professional basketball and football leagues, focusing on five top-level leagues in each sport. The main goal was to quantify home advantage and compare it between the leagues.

Analysis of tourist movements with the help of Geocaching gameDiploma Thesis · 2020

Geocaching is an outdoor game in which players search for hidden caches at outdoor locations with the help of a website and mobile application. Used web scraping to collect the data users enter into the application and then analyzed the collected data.

Basketball and Football dataset2022

Collection of high-granularity data of basketball and football matches obtained using web scraping.

Clustering web users based on mouse movement2021

The objective was to classify web users into groups by performing clustering on their mouse movement data.

Predicting Bundesliga matches outcome2020

Collected data of German Bundesliga matches and used it to build ML models. The goal was to predict match outcomes by creating a three-class classification problem.