Building deployable and updateable deep-learning: the Confused.com journey
This is a story of two worlds, the data science world and the software engineering world. Machine learning is becoming a form of software engineering and machine learning models are becoming so ubiquitous that they are already embedded in most software and services. This is why software engineering practices need to be adopted by data scientists. We’ve seen AI models perform amazing feats in isolation, but how does an established company become data-driven, build a data science practice and harness AI to develop innovative new products and services? Over the course of just two and a half weeks, a traditional business intelligence workflow was transformed into a productionised deep-learning pipeline. In this session we describe how we built and productionised a deep- learning model to predict car insurance premiums, empowering a company to unlock its full potential.