With the LLM work, the readiness of research DSE namespace, and the effort ML team is putting on inference speedup, here are outstanding work for offline inference that need to be captured and worked on:

• be able to deploy batch inference, streaming pipeline
• feature storage on DSE (data as a service concept)