I'm a Researcher in Graph (Network) Data Science, with an educational background in Physics (BA and MA) and a PhD in Network Science.

This means that I take on new problems and develop mathematical and algorithmic solutions for systems that have an interconnected structure (e.g. social networks, biological neural networks). Below are my previous academic and applied projects:

• Publication: Nature Physics | GitHub
  • As a contributing author, developed collision-detection algorithms for $10^{6}$ objects with approximately 3000 unique labels.
  • Detected spatial neighbors by computing pairwise distances of 3D objects, computed with point clouds and kd-trees.
  • Using graph analysis, found a relationship of neuron synaptic connections to the number of their spatial neighbors.

• Publication: EPJ Data Science | GitHub
  • As a contributing author, cleaned and processed the data of 6 timestamped graph datasets (e.g., emails, Twitter).
  • Performed null hypothesis testing for 4 reference null models (randomized timestamps and graph topology).
  • Discovered statistically significant results that Twitter exchanges are less reciprocal and bursty than SMS, calls, and emails.

• Publication: arXiV | GitHub
  • As a leading author, performed a comprehensive data processing and analysis of 15 volumetric 3D graph datasets.
  • Developed algorithms to quantify the shape, size, and geometry of the data (e.g., fractal dimension, edge volume).
  • Created a pipeline that randomizes physical edge trajectories to detect obstacles for more than $10^{5}$ edges at once.

• Status: In progress | GitHub
  • As a leading author, simulating 2D and 3D graphs' physical attacks (spatial edge removal) in their embedding space.
  • Developing a measure to quantify the connectivity of spatial regions in which the graph (network) is embedded.

• Event: Citadel - Correlation One Global PhD Datathon 2023 | Competition Link | GitHub
  • As an individual competitor, utilized NLP methods to determine sentiment, emotion, and topic of text data.
  • Computed correlations of click-through-rates to determine what drives user engagement, with Google Analytics data.
  • Developed a custom ranking of clickbait articles, that relied on their daily, aggregate, and top 5% performance.

• Event: Optiver - Trading at the Close (Kaggle) Competition | Kaggle Link | GitHub
  • As a leader of a 3-member team, implemented an XGBoost model on a data stream to predict stock prices.
  • Optimized hyperparameters using k-fold cross-validation tailored for time-series data, including periodic retraining.
  • Automated data-stream tasks: data collection and cleaning, feature engineering, model retraining, and prediction.

• Organization: WUTIS - Academic Trading And Investment Society | LinkedIn | GitHub
  • As a leader of a 4-member team, achieved a first-place victory in the Algorithmic Trading pitch competition.
  • Created a graph representation based on the cross-correlation of stock price time-series data to identify group trends.
  • Backtested an algorithmic trading strategy based around stocks deviating and returning to group trends in the graph.