Anyone who has built a machine learning model will know the feeling… "How do I get my masterpiece out of this python notebook and in front of the world?". Answering this question is rarely simple and with a multitude of different options to consider, this can be a huge source of both technical debt for data science teams and dependency on engineering resource. At HeadBox we have developed a lean deployment pipeline for simple machine learning models that are used in our venue recommendation engines. Here I will demonstrate the deployment of a simple classification model using three Serverless lambda functions, pulling data from a data warehouse such as Snowflake, posting results to S3 buckets and DynamoDB tables, as well as posting daily performance updates to slack. Our first Serverless function will be used to pull training data from Snowflake, perform feature engineering and train a simple decision tree model.