Posters Galore: 'Deploying and scaling models in Data science'

Panagiotis Kourouklidis
University of York 

Panagiotis Kourouklidis, University of York
Email

In recent years, commoditisation of computing and digital storage resources has allowed businesses to extract insights from the data generated by their operations at a scale not witnessed before. Additionally, thanks to advances in the field of AI, a number of tasks that previously required human labour can now be automated. Of course, human labour is still needed in order to develop the system that is going to complete the aforementioned tasks.

The development of such a system starts with a data scientist producing a model, usually by using machine learning techniques and a training dataset. Unfortunately, the model produced by the data scientist is not the only component needed for a production-ready system that can run reliably. Usually, after the model has been produced it is passed on to a team of software engineers so that they build the rest of the system around it. Components of this kind of system include data gathering, cleaning and transformation as well as model deployment and performance monitoring.

The ongoing goal of the presented research project is to develop a framework that can streamline the development of the above components and can be used by a person not necessarily well versed in software engineering. That will allow data scientists to develop systems that are closer to being production-ready while reducing their reliance on software engineering teams.

View the poster here
View more posters and videos