Presentation of the SFT ML inputs to IT for the RCS-ITC Technical Committee Meeting.
The committee requires that the presentation should cover the following three questions:
- What do you currently do in the field of ML?
- What is planned/prepared for the future in your community?
- What do you want IT to provide to you?
A little bit more context about what you could cover in each question is given here:
- Describe the main ML use cases you currently have in your community/experiment
- Give a short overview of ML activities
- Which ML environment do you currently use?
- Which ML environment would your community prefer for more efficient ML?
- What is the typical size of created/used models?
- What is the typical/largest input data set size for a training?
- Is your training currently IO or GPU performance bound in your given environment?
- How many GPU hours do you estimate your community uses per week?
- What are you missing in the current environment available to you?
- Which of the currently available ML platforms and tools in IT do you know/use ?
Describe (if known) the evolution and future use cases in preparation/planning
- Are there significant changes in the way you need to do ML foreseen?
- Will you need new tools, different platform, hardware etc. ?
What are your key expectations from IT to provide … for example
- A scalable ML hardware platform for training, inference aso.
- A comprehensive toolset & services for ML e.g. model registries aso.
- Dealing with copyright & licences for models & tools
- Enabling access to commercial cloud ML platforms
- Expertise on how to evolve/apply/develop models (various ML techniques, RAG pipelines etc.)
- Expertise in the execution/optimisation ML workflows
- Support in the procurement of community-owned resources
- Platform to run ML driven services like expert systems etc.
Introductory courses on the IT ML platform & tools
Any other ..
