Within our software architecture, we have separated out the concerns of control (implemented using state machine modelling and verification tools from Cocotec), algorithms (in C++) and data management (using a bespoke tool). This has had substantial benefits, especially in definition of the system across software and functional engineers, and testability of the resulting software. However, there are repetitive elements of code, especially around the data services, that are specific to our domain (i.e., we have training examples but they are not externally published). This project will investigate the use of LLM (Large Language Model) technology to automate some of this code generation. Findings, and hopefully a proof-of-concept, will be reported to our architect community. Unless security circumstances change, this will be using an on-prem deployment as our code is both commercially and legally (export control) sensitive.