CMS Data Reduction Scientist

Big Science Sweden; Genf; Vor Ort; CHF 70.000 - 90.000; Vollzeit

Zusammenfassung

Big Science Sweden in Geneva, Switzerland, is seeking a professional in data science to define next-generation data formats and optimize compression methods. You will benchmark techniques and develop workflows to leverage physics performance while managing large datasets.

The ideal candidate will have a Master's degree with relevant experience or a PhD, with no prior CERN contracts. Join us in transforming high-level data processing to ensure breakthrough scientific discoveries.

Qualifikationen

• Professional background in Physics, Applied Physics, or Data Science.
• Master's degree with 2 to 6 years of experience or PhD with up to 3 years of experience.
• No previous CERN fellow or graduate contracts.

Aufgaben

• Define data formats for maximum reduction and physics performance.
• Develop workflows for lossy compression impact quantification.
• Benchmark compression methods measuring CPU/GPU costs and latency.
• Enhance lossless compression via tuning methods and AI techniques.

Kenntnisse

Parallel programming

Data compression

Physics background

Benchmarking techniques

Ausbildung

Master's degree or PhD in Physics, Applied Physics, Data Science

Jobbeschreibung

As part of our commitment to advancing science, the Next-Generation Trigger (NGT ) initiative has launched the Real-time Reconstruction Revolution (R³, "R-cube"). R-cube aims to transform the High-Level Trigger (HLT) farm into a full reconstruction farm capable of processing all input data (up to 750 kHz, ~6 TB/s) with offline-like quality through improved reconstruction code and calibrations. A major challenge is managing the storage of the high-quality data produced by R³.

Your responsibilities:

• Defining next-generation data formats that maximize data reduction while preserving physics performance. Ensure they are parallel-programming friendly (e.g., structure-of-arrays, SoA) and optimized for accelerators (e.g., GPUs).
• Developing a workflow to quantify the impact of lossy compression on physics performance, including clear metrics, reference analyses, and automated regression tests.
• Benchmarking compression/decompression methods by measuring CPU/GPU cost, I/O throughput, memory footprint, and latency under realistic workloads.
• Improving lossless compression of raw data by tuning existing methods and designing new ones that leverage R?-reconstructed objects and AI techniques; where feasible, refactoring raw data structures to enhance compressibility and I/O efficiency.
• You have a professional background in Physics, Applied Physics, Data Science (or a related field) and have either:
  • Master's degree with 2 to 6 years of post‑graduation professional experience;
  • PhD with no more than 3 years of post‑graduation professional experience.
• You have never had a CERN fellow or graduate contract before.