While Training has been the shiny object of AI for the last several years, organizations moving beyond development and actually into production are finding that inference is where their AI aspirations become achievable. Unlike training, where systems could be architected to excel at one specific phase of the journey to AI, inference systems often serve multiple purposes and must be capable of excelling at their “day jobs”, while performing these new AI inference tasks in parallel. They’re also diverse, ranging from edge devices to powerful workstations and servers.
This session will discuss inference considerations organizations should consider when putting AI into production, up to and including:
- Understanding the role inference plays, vs training
- The critical role workstations can play with inference
- How and why organizations increasingly inference against pre-trained models
- Architectural considerations for inference