tommy chu

MSc Data Science at EPFL

msc data science student at EPFL. i'm into HPC, efficient deep learning (model compression, pruning, quantization) and using ML on scientific problems (drug discovery, extreme-weather prediction).

Research

  • T. Chu, A. Kovalenko. Projective Pruning by Decoupling Weights. Machine Learning and Knowledge Discovery in Databases (ECML PKDD 2025), Research Track. Lecture Notes in Computer Science, vol 16018. Springer.
    doi.org/10.1007/978-3-032-06106-5_19
  • T. Chu. Machine Learning Techniques for Weather Events Forecasting. Bachelor's thesis, Faculty of Information Technology, Czech Technical University in Prague, 2025. Advisor: A. Kovalenko.
    hdl.handle.net/10467/124324

Experience

  • Student Researcher, Czech Technical University
    • neural-network compression for LLMs (pruning, sparsification, and low-rank weight approximation).
    • spatiotemporal modeling with physics-informed neural networks for hyper-local extreme-weather prediction.
  • Scientific Software Engineer, Merck & Co. (MSD)
    • built data infrastructure for generative models in drug discovery over multimodal molecular representations like SMILES, voxel, point-cloud, and graph, plus a petabyte-scale platform for virtual screening, next-generation sequencing, and combinatorial chemistry.
    • earlier, sped up protein-ligand molecular-dynamics simulations with IO-aware data manipulation and an on-prem HPC to cloud migration.

Education

  • EPFL, MSc Data Science
  • Czech Technical University in Prague, BSc Computer Science

    GPA 4.00/4.00

  • prg.ai Minor, Czech AI research initiative

    interdisciplinary AI program run jointly by CTU and Charles University

  • IBM Quantum Summer School

    qiskit: theory to implementation

Projects

a pile of random things i built for fun over the years, mostly to learn something or mess around. browse them all.