Machine Learning Engineer and Developer in Vienna, Austria

The competition focused on building a machine learning model to speed up disease detection. We worked with thousands of images collected in rural areas to help identify diabetic retinopathy automatically using deep learning.

Highlights:
• It was a medical image classification task.
• I ranked 62nd out of 2943 competing teams (top 3%).
• I achieved a Kaggle silver medal and competitions expert status.

The goal of the project was to analyze customer behavior, identify the most valuable customer segments, and design a sales funnel that is significantly more effective than the old one.

I worked with a product manager to make data-driven decisions and optimize information gains at each step of the funnel. I also optimized ad feedback to FB to enable faster and more accurate learning.

As a result, customer sign-ups tripled in just two months while maintaining the same level of conversion cost.

I analyzed data from multiple-choice tests to reduce testing time and improve the test-taker experience. My work included implementing a correlation analysis for test questions to avoid similar, duplicate, or erroneous questions. I also analyzed information gain per question to remove "low information" questions.

Built quantitative models for predicting stock price movements. Made it to the top 1% of quants out of over 10,000 at Numerai—a hedge fund that crowdsources investment models. It's organized as a tournament, claiming to be "the hardest data science tournament on the planet."

A competition organized by the National Bank of Hungary. The objective is to predict the total export production of Hungarian companies with anonymized data provided by the Hungarian National Bank. Competitors need to predict next year's export income for each company using machine learning tools based on historical data and other statistics.