Title: Experienced Deep Learning Researcher Needed to refine a minor Medical Imaging Project for Journal Publication Job Description: Seeking a highly qualified deep learning researcher/developer to critically review, enhance, and optimize the code for our project to meet the standards of a peer-reviewed journal publication. Our work focuses on skin cancer classification using deep learning with the HAM10000 dataset, comparing three architectures: Vision Transformer, Swin-T, and CoAtNet. While the core implementation is complete, I need an expert to address potential shortcomings and improve reproducibility to strengthen the manuscript. Project Scope & Key Responsibilities: 1. Code Review & Best Practices: Audit existing code (Python/TensorFlow/Keras) for technical flaws, efficiency, and adherence to reproducibility standards. 2. Methodology Validation: Verify statistical significance of results. Check evaluation metrics (e.g., AUC-ROC, sensitivity, specificity, F1-score) and ensure comparisons are fair (same training conditions, hardware, preprocessing). 3. Model Improvements: Suggest architectural tweaks (e.g., attention mechanisms, hybrid models), hyperparameter tuning strategies, or regularization techniques to boost performance. Explore explainability methods (e.g., Grad-CAM, SHAP) to strengthen clinical relevance. 4. Reproducibility & Documentation: Ensure code is fully reproducible with clear READMEs, environment files (Docker/conda), and preprocessing steps. Assist in creating figures/tables for the manuscript (e.g., confusion matrices, ROC curves, feature visualizations). 5. Ideal Candidate Qualifications: Hands-on experience with HAM10000 or similar dermatology datasets (e.g., ISIC Archive). Strong familiarity with ViT, Swin-T, CoAtNet, and their implementations in TensorFlow. Experience preparing code for academic reproducibility (e.g. GitHub templates). Ability to communicate technical feedback clearly and collaborate on manuscript revisions. 6. Deliverables: A detailed report of code improvements, methodological gaps, and performance enhancement strategies. Refactored, well-documented code with reproducibility checks. Draft Manuscript and Assistance in responding to reviewer comments post-submission.