ML Operationalization: From What? & Why? to How? & Who?

Abstract: If machine learning (ML) is as valuable as we think, why do only a sliver of projects make it to production or value realization? Companies that have dared to implement ML in production have quickly realized there’s a huge gap between the decision to implement and the actual execution. More often than not, businesses on their ML journey are finding that scaling from a handful ML services to the hundreds ML Applications expected by the business, demands processes that go far beyond the typical data science and DevOps workflows. This is because Data scientists, who are tasked with this delivery, while being very well versed on the What and Why of ML from a business and data science perspective, are typically lacking the experience or expertise of tackling How it will actually work, and Who needs to work together to get things done. And by the way – expecting this from the data teams only leads to frustration on every side of the equation, and to very little tangible results.
As business needs for ML-based applications surge, and the gap continues to grow, a new paradigm – MLOps – is emerging. MLOps enables enterprises to establish an automated and scalable process, where ML investments are able to generate business benefits across the entire enterprise, and MLOps teams, whose who inevitably own the service delivery, can emerge both successful and accountable. By creating a more holistic approach to operationalizing ML, enterprises can more rapidly include ML in vertical and horizontal applications, build more intelligent applications, and form the structure required for an ML-driven business – all while enabling collaboration among relevant stakeholders across the company. In this talk, we will discuss how leading-edge enterprises are bringing together the people, processes and technology to see ROI from ML.

Bio: Sivan brings close to 20 years of experience in enterprise software, and leverages his background to help companies successfully take Machine Learning into Production with ParallelM. Prior to ParallelM, he founded and served as GM of two business units, Kenshoo Social and Kenshoo Local, at Kenshoo, Inc., an enterprise software company for leading online marketers. Prior to Kenshoo, Sivan spent close to a decade at Mercury Interactive, prior to its acquisition by HP in 2006, where he held numerous business and product leadership roles across the company.

Open Data Science Conference