Our project aims to develop an advanced review tool leveraging LLMs to identify and reduce biases in medical documents. The platform allows users to upload medical documents, from which text is extracted and, after some preprocessing, analyzed by six specialized bias detection models. Sentences flagged for potential bias are presented to reviewers alongside context, enabling users to validate the results. Correctly identified biases are stored as positive examples with user explanations; incorrect ones are stored as negative samples. All labeled data contribute to ongoing model fine-tuning, creating a feedback loop that enhances bias detection accuracy over time.