This week's video covers the biggest challenge developers experience when deploying LLM-based solutions 👉 getting visibility into how well they are working.

We explain this in Prolego's Performance-Driven Development (PDD) methodology, a framework for building LLM-based solutions. In short, you can achieve this visibility using a Performance Evaluation Framework that consists of four key components:

1. A representative set of data and tasks for the problem you're trying to solve.
2. The AI solution, including prompts, data preprocessing, custom libraries, and prompt orchestration.
3. The evaluation workflow, which runs your data and tasks through the AI solution.
4. A performance report listing expected versus actual results for the set of tasks.

A simple spreadsheet listing tasks and expected outcomes is enough to get started. For a full explanation of PDD, check out Prolego's GitHub repository.