Computer Science · Machine Learning

Marko Petrov

BSc. Computer Science · Skopje, North Macedonia

I work on competitive machine learning, with a focus on validation, feature engineering, and building reliable end-to-end systems.

Interests: machine learning research · computer vision · algorithms

Currently working toward Kaggle Competitions Master.

Monochrome engraved illustration of a cow overlooking a mountain lake at sunset
Figure 1. Perspective, persistence, and the long view. Current view: Original

Abstract

I am a third-year student at , focused on machine learning research and competitive ML.

My work spans tabular forecasting, 3D computer vision, and algorithmic problem-solving. I care most about sound validation, predictive features, and systems that hold up beyond a public leaderboard.

1Education

Faculty of Computer Science and Engineering

Ss. Cyril and Methodius University · Skopje

BSc. Computer Science

2024–2028
expected

GPA 9.62/10 · Dean’s List · Full scholarship for competitive-programming achievements

2Kaggle Competitions

is the world’s largest machine-learning community, with more than 25 million registered members. Its hidden-test competitions attract top practitioners; each project below represents roughly two months of focused work.

  1. [1]

    Basketball Tournament Prediction

    Silver Medal · Top 2%

    Built an ensemble for Kaggle’s NCAA March Madness competition, using season-level features and out-of-fold validation across held-out seasons to predict matchup point differentials. Calibrated margins into win probabilities with spline interpolation and optimized for Brier score.

    Strong validation and feature engineering drove the largest gains · approximately 3,600 competitors

  2. [2]

    3D Cell Tracking During Zebrafish Development

    Top 12%

    Built an end-to-end pipeline for anisotropic 3D time-lapse microscopy: cell-center detection, temporal matching, division prediction, and lineage reconstruction. Combined learned modeling with tuned post-processing in a reproducible inference notebook.

    First deep-learning competition · approximately 4,000 competitors

3Competitive Programming

is algorithmic problem-solving under strict time and memory limits, emphasizing precise reasoning, data structures, and efficient implementation.

  1. 2023

    International Olympiad in Informatics

    Honourable Mention. Represented North Macedonia at the 35th IOI in Szeged, Hungary, after only one year of competitive-programming experience. The IOI is one of five international science olympiads; countries typically send four students to compete individually over two days.

  2. BOI

    Balkan Olympiad in Informatics

    Competed for North Macedonia at the regional olympiad.

  3. MOI

    Macedonian Olympiad in Informatics

    2nd place overall.

End note

See you at the summit.