Developed and integrated the LlamaCppEmbeddings feature within the Embeddings module of the argilla-io/distilabel repository, enabling support for loading Llama.cpp models from both local file paths and the Hugging Face Hub. Leveraged Python and machine learning techniques to implement GPU acceleration and normalization, enhancing offline capabilities and reducing inference latency. Comprehensive unit tests were added to ensure the reliability of embedding generation, including coverage for edge cases. This work improved model deployment flexibility and robustness, allowing users to efficiently generate embeddings in diverse environments while maintaining code quality through systematic testing and adherence to best practices in LLM integration.