Machine Learning DevOps Solution

Using Machine Learning in productive applications requires certain qualities from software development, at the same time ML requires a vast number of experiments to tune hyperparameter try different model architectures. Many ML applications required a continuous update of the model based on changing or extended data requiring ML DevOps.
Therefore, we have developed a Python framework building on established tools, which helps ML developers to focus on data centric model development and testing by triggering tools required for a systematic development as far as possible automatically in the background.
We show how our framework orchestrates several tools in a typical ML application and how our framework supports the whole ML training process and deployment.

Vorkenntnisse

Experience in machine learning and knowledge of the typical difficulties in machine learning software development (like making machine learning reproducible, continuously deploy trained models)

Lernziele

* DevOps challenges typically faced in machine learning development
* Which possible solutions exist to manage potentially large amount of model and data variants
* Which open source tools exist and how could they be orchestrated

 

Speaker

 


Sven Pelzer ist Softwareerchitekt, Entwickler, DevOps. Er arbeitet seit 20 Jahren als Softwareentwickler (JAVA, .NET, Python).


Patrick Levi is Physicist, data scientist, software developer with a focus on machine learning software solutions.

enterPy-Newsletter

Sie möchten über die enterPy
auf dem Laufenden gehalten werden?

 

Anmelden