Data/ML/AIEmergingNewAround since 2023
DSPy / Prompt Optimization
Also written as DSPy, Prompt Optimization, Automatic Prompt Optimization
Improving prompts automatically instead of by trial and error: the team defines the task and a way to score results, and software searches for the prompt wording and examples that score best. DSPy, from Stanford, is the best-known framework.
Think of it like
Letting a sat-nav find the fastest route by testing them, instead of guessing from memory.
Junior or senior?
Signals an evaluation-driven engineer.
Senior sounds like
Can describe the scoring method they optimised against and a case where the optimised prompt beat a hand-written one (or didn't).
Ask them
“What were you scoring the prompts on, and how much better did the optimised version do in real use?”