Developed and integrated a Steered Transformer capability across the red-hat-data-services/lm-evaluation-harness and swiss-ai/lm-evaluation-harness repositories, enabling fine-grained control over Hugging Face model behavior for NLP evaluation tasks. This work introduced a new 'steered' model type that loads steering configurations from PyTorch files or CSVs, supporting rapid experimentation and business-aligned model personalization. The implementation included updates to documentation and expanded test coverage to ensure reliability and cross-repo consistency. Leveraging Python, PyTorch, and CI/CD practices, the developer established a foundation for scalable model steering, facilitating faster feature iteration and broader applicability in machine learning workflows.