Conducted during a six-month internship at Amadeus IT Group, a leader in the travel economy, this thesis explores the deployment of machine learning models in cloud environments, specifically Microsoft Azure. It addresses challenges in serialization and analyzes different formats, including PMML, ONNX, and MLflow, comparing their key strengths and limitations. It also proposes solutions for deploying models as cloud microservices, highlighting potential improvements in flexibility and efficiency compared to an in-house backend deployment.