Research

Spotify

Currently, my research focuses mainly on user retention prediction and understanding, attribution modeling, and time series clustering.

I have been (co)supervising MSc and PhD students for various research projects:

  • Stefan Magureanu, Optimizing Search Result Ranking (PhD internship)
  • Foivos Michelinakis, Passive Mobile Bandwidth Classification Using Short Lived TCP Connections (PhD internship)
  • Gustav Mörtberg, Detecting Frauduent Users (MSc internship)
  • Martin Håstad, Evaluating Sessionization Methods in Music Streaming Context (MSc thesis)
  • Erik Aalto, Learning Playlist Representations for Automatic Playlist Generation (MSc thesis)
  • Theodoros Vasiloudis, Extending Recommendation Algorithms by Modeling User Context (MSc thesis)

Delft University of Technology

I studied the system dynamics and user behaviors in Peer-to-Peer (P2P) systems like BitTorrent, through extensive measurement and statistical modeling. I was also one of the main contributors of the Peer-to-Peer Trace Archive — one of the largest public archive for P2P traces.