Wanna know how to run Tensorflow in production effectively? How to customize your input pipeline?

In this event we will have three experienced Tensorflow Engineers sharing their insights. The event will take place at Omio, (thanks for hosting us )

Come and join for the tensorflow night!

Agenda:

6:45 - 7:00 pm: Welcome to the meetup, grab some food and drinks

7:00 - 7:20 pm: Tensorflow Serving - Brecht Coghe

7:30 - 7:50 pm: Tensorflow Dataset api - Victor Sonck

8:00 - 8:20 pm: Tensorflow Records - Ron Hagensieker

From 8:30 pm: Networking, stay for some drinks

TF Serving
Tensorflow is nice for doing research, however, one of the main advantages of the Tensorflow framework over other ML frameworks is it's ecosystem that easily allows for running your code in production. That is why we want to show you how to make use of tf serving.

TF Dataset api
Most of the Tf users are familiar with creating models and running training by now. What people are more unaware of is the tf dataset api. With this module, tf allows for preprocessing, preloading and customizing your input pipeline for training your models.

TF Records
Tensorflow's recommended way to store data on disk is as binary TFRecords file. This is a logical data source to transfer data into Tf's own Dataset objects and especially handy for storing training data. Best practices and recommended workflows are presented in this talk to make most use of this file format.

Brecht Coghe
After his master in AI and being a PhD candidate at the KU Leuven, Brecht started working as an ML Engineer at ML6. He works on multiple projects creating insights for the customer and building intelligent solutions that tackle the customer’s immediate business needs. Next to these projects, Brecht is also active in Applied Machine Learning research that focuses on improving current logistic tools.

Victor Sonck
Victor is a machine learning engineer working mainly on computer vision and image recognition, as well as getting these solutions to work as fast as possible on the edge. Additionally, he's dabbled in genetic algorithms and reinforcement learning.

Ron Hagensieker
A learned geographer, Ron entered the domain of machine learning through image recognition tasks in satellite remote sensing. After finishing his PhD thesis on the classification of deforestation sites, he worked in positions as remote sensing engineer and data science consultant before starting at ML6. There he joined as machine learning engineer at the recently founded branch in Berlin.