"I was a happy data scientist until we decided it was time for deploying our models." It is common among many DS/ML teams that when the time for productionizing the model comes, they are caught off guard due to poor planning. Of course, thinking solely about the end is far from enough, the stages beforehand are equally as important. To reach the end of any endeavor we need to be strategic, the same applies for succeeding with MLOps. One such strategy is to take on a less challenging problem or part of it in the beginning and find the easiest way it can be solved.

We live in an age of rapid AI innovation and progress. Yet even as academics and researchers make astonishing advancements, demonstrating real business value and positive return on investment is challenging. Developing cutting edge AI applications based on machine learning models integrated with existing business software is a common challenge. This article discusses a few of the core pain points and strategies to address them. The first challenge most organizations encounter is the increased complexity of preparing data and dataset management.

In part 1 of this article, we observed that due to a lack of a standard definition and adoption of AI productionization, there is disillusionment amongst businesses on achieving true impact of AI and Machine Learning. This not only can hold the field back from achieving its true potential, it risks marring the reputation of emerging technologies in the eye of executives, potential beneficiaries and the general public. So what are some potential issues and resolutions to tackle productionization? Tl/dr is towards the bottom if you want to skip there. It may disappoint tech enthusiasts but imho, very little depends on tooling itself.

For those who are involved in AI and Machine Learning, may have noticed a trend recently. While new announcements in algorithmic advances like GPT-2 from OpenAI, EvaNet from Google and considerably more that I can't list here, have all garnered interest, industry at large seems to feel something is missing. That missing feeling is the dirty P word - productionization. This is troubling, not just for the practitioners but for the field as a whole. Before we go further in diagnosing the cause and identify remedies, let's put a definition on what productionization is (Disclaimer: this is my perspective, feel free to add a comment).

Despite incredible recent advances in machine learning, building machine learning applications remains prohibitively time-consuming and expensive for all but the best-trained, best-funded engineering organizations. This expense comes not from a need for new and improved statistical models but instead from a lack of systems and tools for supporting end-to-end machine learning application development, from data preparation and labeling to productionization and monitoring. In this document, we outline opportunities for infrastructure supporting usable, end-to-end machine learning applications in the context of the nascent DAWN (Data Analytics for What's Next) project at Stanford.