COOP is part of the CERFACS COOP-Algo Team since Jan 1st 2020. The team encompasses a wide array of activities, including Computer Science and Engineering, High Performance Computing and Data Science for Computational Physics.
It is focused on bringing support to all research teams such as tools, best practices and resources to produce quality high performing scientific software, and technology survey to take advantage of new approaches at the right time. The missions of the team involve:
- Code hardening:
- API design
- Graphical User Interface design
- code quality (a.k.a. linting)
- unit testing and code coverage
- code robustness and maintainability improvements
- continuous integration and deployment
- High Performance Computing:
- parallel optimization
- CPU/GPU/TPU computing
- Technology survey:
- next generation architectures
- machine and deep learning
- quantum computing
If you wish to find out about opportunities to work in the team, please visit the CERFACS open positions catalog.
Due to the Covid-19, we moved to a text-oriented approach of the work. This is well described in the Basecamp’s “How we communicate”. This is also why we have this blog running. By reading it, you will grasp a fair part of what we are doing here.
A part our production is open-source, mirrored in the Cerfacs Gitlab.com. We strongly rely on OpenSource developments:
- Easier to find national or European funds.
- Good for our non-permanent members who can show off during job interviews.
- Push to higher quality. See our post on the OpenSource gold standard for python packages for example.
- Loads of open source-only great services : gitlab.com, PyPI, Readthedocs.
- Confidentiality (yes!): by splitting our devs into OpenSource and Private repositories, everyone in our staff is aware of what is confidential and whats not.
- Facilitates collaboration, because anyone can see and use the same code without restrictions.
There are also private repositories dedicated for our customers, with the same level of quality as the open-source production.
Disclaimer: our open-source softwares are provided “as is”. By using them users accept that it may contain defects or not meet their expectations.
Gabriel Staffelbach is an expert in high performance computing (HPC), among other scientific calculus challenges.
Antoine Dauptainimprovement of research software at CERFACS, from early proof of concepts to fully mature softwares.
Corentin Lapeyre currently focuses on machine learning (and specifically deep learning) applications for physical modeling.