Sergei Smorodskii developed a score-aware embedding retriever for the stanfordnlp/dspy repository, introducing the EmbeddingsWithScores feature to expose similarity scores alongside retrieved passages. Using Python and leveraging skills in data retrieval and machine learning, Sergei refactored the retrieval pipeline to handle score computation within the forward method and improved code maintainability by adding strict zip handling and comprehensive docstrings. This work enabled downstream components to filter and re-rank search results based on relevance scores, supporting more reliable confidence calibration. The changes enhanced interface stability and documentation, addressing maintainability while directly improving the quality of search and retrieval workflows.