Analysis from McKinsey where basically they canvassed variety of use cases across industries and looked at what was the incremental sort of thing you had and what they found is that the latest generation of techniques around deep learning reinforcement learning etc on average had about a 62% uplift in value which is really staggering and obviously is this significant amount.

Of investment they’ve therefore going.

Into R&D around data science and this mean the millions of models will be produced over the next few years.

You all know the S that any new technology that is created generally follows this kind of test curve where you have this kind of a period of deep tech technology.

Adoption and the kind of levels out at the top and what machine learning and AI is enabling is this kind of continuing acceleration from an s-curve to more of a sort of J curve as more and more processes.

Get automated and you know that really is you know extremely game-changing for all industries so a lot of people focus on the role of the data scientist and you know they.

Obviously play a critical role but I think what is often underappreciated is the importance of DevOps in deploying machine learning and that’s really kind of the main focus of this talk obviously business student users as well are you know finding the problems to solve issuing budget and influencing the business KPIs but.

Really the two technical roles here need to kind of work in sync in order to put machine learning into production and actually the data science and the and the.

Do completely do completely different things so they decide to star in charge of developing models and data engineers are often closely connected are creating the pipelines to feed that data into the data scientists and they speak a completely different language to the DevOps team who actually responsible for production izing those models and enabling them to actually run within your business applications and.

What we find is this often kind of like a in a silo these teams are often siloed especially out within larger organizations you know they’re not sitting next to each other they’re in different different rooms different departments and they sister you know causes poor feedback.

Loop and there’s often quite a long delay between data scientists creating a model in actually being productionize and there’s a third role which is seeing emerging now and sort of trend emerging around this concept of ML ops and EMA ops effect to be either kind of into the interdisciplinary skill set which combines understanding of data science and DevOps which is really hard to find so what we’re.

Producing is and what lot of people in the industry of producing are technologies that can enable data scientists to tackle more DevOps related challenges and vice-versa so there are two broad workflows in data science one of them is model development and the other one is a model serving and this is quite a simplistic pipeline here you have you know training data cleaning at the top moving through to feature engineering and building models once the.

Model is being created you have a kind of live pipeline which is which sometimes.

LEAVE A REPLY

Please enter your comment!
Please enter your name here