[Pycon] [new paper] "Christian Barra" - Let’s talk about MLOps

[info a pycon.it]

Title: Let’s talk about MLOps
Duration: 45 (includes Q&A)
Q&A Session: 0
Language: en
Type: Talk

Abstract: The quantity of hype around machine learning and AI is probably second only to bitcoins and blockchains.

But until a machine learning model is deployed to production the value delivered to companies is approximately equal to zero.

Together with the common mantra that data science can’t use agile/lean frameworks or that the best software engineering practises don’t apply explains a lot about why often companies got burnt with their data science projects and why generally they under delivered.

MLOps is here to help, the machine-learning equivalent of DevOps: it solves the problems of implementing machine-learning in production.

During this talk I will introduce the data science lifecycle, the concept of machine learning Ops, its characteristics, why is extremely required, how it compares to DevOps, how it will become a required capabilities for any DS/ML/AI Team, tools available and how you can start with it.

Tags: [u'machine-learnig', u'agile', u'dataops', u'devops', u'lean', u'datascience', u'Artificial Intelligence', u'pydata']