The development life-cycle of intelligent applications has witnessed a significant change with the advent of generative AI. This is indeed a paradigm shift from classical ML - where we train a model with specific statical algorithms to deliver pre-defined outcomes - to LLMs based solution - where we add our own data to pre-trained LLMs to get a non-deterministic outcome. Such a change has larger implications not only on how we build applications but also in how we test, evaluate, deploy and monitor them – leading to the introduction of LLMOps.
In this session we are going to explore the foundations of LLMOps, providing examples of tools and practices at our disposal in the Azure ecosystem.