At the core of Teradyne’s Analytic Management Platform (AMP) is the UltraEdge2000 server, capable of executing Machine Learning models during device test. When moving any software solution from development to production deployment, real world possibilities must be considered. This paper will discuss lessons learned and best practices to be followed when developing and deploying machine learning models on UltraEdge2000 from an IG-XL test program. Topics to be covered include – Deployment methods – Docker or XAR, which is best and when? Communication methods – Sockets or FIFOs? Adapting to testers that do not have UltraEdge hardware or libraries Best practices for server applications developers.