Prompt engineering

At the start of this month I accepted an opportunity to step away from my usual team and join a short-term AI project for the company. My role involves supporting data scientists by productionising their code and building backend pipelines.

This is my first time working with data scientists using LLMs. They attempt to guide the LLM using carefully worded prompts, apparently leaning on trial and error to tease out desired responses. Essentially, the LLM is a black box. The lack of determinism isn’t my cup of tea, but I’m happy for others who enjoy this prompt engineering.