Data Reduction Physicist (EP-CMS-TDQ-2026-146-GRAP)

<p>The NGT Real-time Reconstruction Revolution will transform the CMS High-Level Trigger into an offline-quality reconstruction facility processing up to 750 kHz of events. To make this sustainable, you will lead the design of two complementary data tiers: a low-level reconstructed format that enables trigger rates several times higher than conventional raw-data output while preserving the the option for future reprocessing, and a compact analysis format to save the full 750 kHz stream of online reconstructed events.</p><p><strong>Your responsibilities:</strong></p><ul><li>Design next-generation data structures that push the limits of data reduction to save 750 kHz while preserving physics performance and analysis flexibility.</li><li>Ensure all formats are accelerator-native (e.g., structure-of-arrays, SoA) and optimised for high-throughput GPU processing.</li><li>Build an end-to-end framework to rigorously quantify the impact of lossy compression, with clear metrics, reference analyses, and automated regression tests.</li><li>Benchmark compression/decompression under realistic workloads: CPU/GPU cost, I/O throughput, memory footprint, and latency.</li><li>Advance lossless compression, leveraging R³-reconstructed objects and pioneering AI/ML techniques.</li></ul><p><strong>Your profile:</strong></p><ul><li>Demonstrated contributions to trigger and/or reconstruction in HEP (or comparable high-throughput scientific software).</li><li>Practical understanding of of end-to-end HEP experiment operations, from detector readout to reconstruction, calibrations, datasets, and final physics results.</li><li>Experience working in a large international collaboration (code review, CI/CD, documentation) is a plus.</li><li>Knowledge with LHC experiments and their data formats is a plus.</li><li>Expertise in data compression techniques is a plus. (lossless and/or lossy).</li><li>Experience applying AI/ML methods (e.g., autoencoders) to data reduction is a plus.</li><li>Proficiency in GPU programming and heterogeneous computing is a plus.</li></ul><p><u>Skills:</u></p><ul><li>High proficiency in C++, Python, and ROOT.</li><li>Solid understanding of event reconstruction, including calibrations and commonly used data formats in HEP.</li><li>Spoken and written English, with a commitment to learn French.</li></ul><p><u>Eligibility criteria:</u></p><ul><li>You are a national of a <a href="https://home.cern/about/member-states" rel="noopener noreferrer">CERN Member or Associate Member State</a>.</li><li>You have a professional background in Physics (or a related field) and have either:<ul><li>a <strong>Master's degree with 2 to 6 years</strong> of post-graduation professional experience;</li><li>or a <strong>PhD with no more than 3 years</strong> of post-graduation professional experience.</li></ul></li><li>You have never had a CERN fellow or graduate contract before.</li></ul>