I’m developing an academic-grade system that flags fraudulent credit card transactions and need a skilled ML partner to push it across the finish line. The core pipeline is already sketched out—data ingestion, preprocessing, feature engineering, and model training in Python—but I want it sharpened for the highest possible prediction accuracy. Here’s what I still need: • Clean, reproducible code (Jupyter Notebook or .py modules) that takes raw transaction data through preprocessing and balanced sampling techniques. • Rigorous feature selection or creation that genuinely boosts performance while keeping the workflow interpretable. • Comparative training of several classifiers—think Gradient Boosting, XGBoost, LightGBM, or any ensemble you trust—followed by a clear evaluation with ROC-AUC, precision-recall, and an explanation of any trade-offs. • A concise report (couple of pages is fine) summarising methodology, achieved metrics, and recommendations for further improvement. High prediction accuracy is the top priority; lowering false positives is useful but comes second. If you have experience building similar fraud or anomaly-detection models, especially on credit card datasets, highlight that in your message and point me to any public repos or papers you can share. I’m ready to start as soon as I find the right fit and will be available for quick feedback loops throughout the job.