Towards MLOps in a startup company

Companies which have developed around a product or service built on top of one or several machine learning applications will reach a point where the complexity of these applications become unfeasible to manage manually. To approach a more mature level of their machine learning applications these companies must start implementing MLOps in their operations as well as orchestration tooling for building durable and scalable machine learning pipelines. This paper presents how a startup company in the marketing- and market research sector has taken their first steps towards machine learning maturity by implementing MLOps practices and orchestration with Apache Airflow.