CERN — IceCube Neutrino Observatory

Working with CERN (the European Organization for Nuclear Research) in the context of the IceCube Neutrino Observatory — a cubic-kilometer particle detector embedded in the Antarctic ice — I built high-performance software for one of the most data-intensive experiments in physics. IceCube records the faint Cherenkov light left by neutrino interactions, and separating true events from detector noise is a fundamental signal-processing challenge.

I implemented machine-learning-based noise-reduction algorithms in C++ for neutrino event detection, improving classification accuracy by 28% and reducing false triggers in the reconstruction pipeline. Performance mattered: the algorithms had to run efficiently over enormous event volumes without sacrificing numerical stability.

I also built automated calibration and data-validation tooling for the muon detectors, processing 100GB+ of raw detector data every day, and optimized high-volume distributed data pipelines for throughput and numerical stability in a scientific-computing environment. The work combined low-level C++ optimization with statistical machine learning and large-scale data engineering.