CROATIA

Data Science Economy 2020: Data Vs Covid
9 talks / 4 workshops / 1 panel
delivered from experts & practitioners

15th & 19th of June 2019, Online
https://i2.wp.com/ds2economy.com/wp-content/uploads/2020/06/Martin-Freser-w.jpg?fit=320%2C246
Data Scientist, Adacta (BE-Terna), Slovenia

Martin Frešer


BIO:

Martin’s main interest in machine-learning is working with time-series structured data, but also tackles other fields, such as optimisation and recommender systems. After master graduation at the University of Ljubljana from Mathematics and Computer science he joined Institute of Jožef Stefan in Ljubljana, where he developed his passion to solve seemingly un-realistic problems.  He worked in the artificial intelligence lab, where the main focus was on data mining and machine learning from various signals (sound signals, ECG, environmental signals, …).

From 2015 he joined Data science team at Adacta, where knowledge from academia and research is applied to real-world business problems. With projects at numerous fields, such as manufacturing, retail, digital marketing, energy industry, stock optimisation, the team is enriching business partners with insights, that are hidden in their huge amount of data.

WORKSHOP

Machine learning in Python with serverless Microsoft Azure Functions

Successful machine learning projects often starts with a proof-of-concept implementation and experiments since the customers need to be confident that the solution will fit into their workflow and that it will perform as expected. We, as machine-learning solution providers have to provide a holistic overview of the implementation path, end results and most importantly the costs. To estimate each of the listed points we must evaluate the real-time performance of the solution. It usually occurs that the customer is not willing to invest into hardware before knowing that the ML solution will bring some benefits, so we have to take the advantage of the cloud technology to produce preliminary results in fast and cheap manner. Moreover, we can utilize the visualization tools to provide end results in more intuitive fashion, so we can bridge the gap between understanding complex machine-learning process to non-machine-learning audience.

In this workshop, we will show how to connect your machine-learning project pipeline that runs on on-premises, to server-less service called Microsoft Azure Functions. We will use Python and other open-source libraries like scikit-learn and pandas to implement the connection and deploy it to Microsoft Azure. The interpretation of the results is done in Power BI. Once the service is created, we can provide the solution for testing and evaluation to any customer.