Offre en lien avec l’Action/le Réseau : BigData4Astro/– — –
Laboratoire/Entreprise : Institut d’astrophysique de Paris
Durée : permanent
Contact : dubois@iap.fr
Date limite de publication : 2024-07-31
Contexte :
The Institut d’Astrophysique de Paris (IAP) is a Joint Research Unit of CNRS and Sorbonne University, with approximately 150 researchers, post-docs, PhD students, engineers, technicians, and administrative staff. Large-scale numerical analysis holds a significant place at IAP: processing massive astrophysical data (HPDA) and conducting and exploiting numerical simulations (HPC). IAP hosts dedicated computing farms for these applications, the most recent specialising in CPU-GPGPU co-computation and AI use. Among the 80 regular users of these machines, half are collaborators from other laboratories in France and internationally.
IAP is heavily involved in major ground-based and space astronomical observation programs (Euclid, JWST, SVOM, LISA, Ariel, GRAND) and at the forefront of numerical modeling in astrophysics with the most extensive simulations in the field (Horizon, Sibelius), conducted on national (Joliot Curie, Jean Zay) and international (Nurion, Cosma) supercomputers. IAP is a leading center for innovative statistical analysis of large data volumes (BORG, Learning-the-Universe projects). These commitments entail new development and deployment needs for codes, including optimization for GPGPU architectures and the use of AI tools.
As the key person for code and algorithm development and optimization on HPDA/HPC machines for new GPGPU architectures, the research engineer will join the “technical support for projects” team of 14. This activity will enhance the laboratory’s capability and expertise in HPC/HPDA themes. (edited)
Sujet :
### Role:
The research engineer will provide expertise in the use of GPGPUs for high-performance data analysis (HPDA/HPC) within research projects. They will contribute to solving theoretical astrophysics problems through numerical simulations and observational astrophysics through the processing and analysis of large data sets.
### Activities:
– Develop and optimize HPC/HPDA numerical codes, particularly those based on AI, for large data sets and numerical simulations.
– Develop and port codes to GPGPUs, including on distributed memory parallel computing machines, considering specific architectures.
– Monitor technological advancements in computational libraries and new statistical methods (AI in a broad sense).
– Deploy codes on national and international supercomputers (tier-1 and tier-0).
– Draft technical information for computing and storage time requests.
– Participate in data management (transferring input data to external machines, retrieving data, and organizing long-term storage).
– Define the hardware and software infrastructure for the unit’s HPC/HPDA machines, interact with service providers, and co-author procurement requests.
Profil du candidat :
Formation et compétences requises :
#### Knowledge:
– Algorithms
– Scientific computing
– Statistical methods
– Computational libraries
– Compiled languages (such as C, C++, Fortran)
– UNIX-family operating systems (such as Linux, BSD)
– English language: B2 to C1 (Common European Framework of Reference for Languages)
– Scripting languages for scientific computing (such as Python, Julia, Matlab)
– General scripting languages (such as shell, Perl)
#### Operational Skills:
– High-performance programming, particularly in GPGPU
– Use of source code management tools (such as GitLab, GitHub, Bitbucket)
– Technical document writing
– Reporting activities
– Technological and bibliographic monitoring
#### Personal Qualities:
– Ability to work independently
– Strong interpersonal skills
– Ability to work with multiple stakeholders
Adresse d’emploi :
98 bis boulevard Arago, 75014, Paris