I have already fabricated both a conventional and a modified CCO, measured their performances, and logged every parameter in one consolidated Excel sheet. What I still need is a Python-based machine-learning workflow that learns from those results, predicts the required transistor W/L ratio for any new specification, and then surfaces the best-performing, fully optimised design point. The dataset is cleanly organised in Excel columns, but it still needs to be split into training, validation, and test sets, with any necessary feature engineering you judge appropriate. I have no fixed preference on the final algorithm—linear models, tree ensembles, or a small neural network are all acceptable as long as they deliver solid predictive accuracy and are easy to retrain when I add more data. Please build the solution in standard Python tooling (pandas, scikit-learn, TensorFlow or PyTorch only if the accuracy gains justify it) and present the work in a Jupyter Notebook. Your notebook should walk me through: • data import, preprocessing, and exploratory visuals • model selection and cross-validated performance metrics • prediction of W/L ratio on unseen inputs • a short optimisation routine that searches the design space and highlights the top candidate settings based on the trained model’s outputs Alongside the notebook, include the cleaned CSV you actually trained on, any helper scripts, and a one-page summary that explains model choice, key findings, and how I can rerun the pipeline when fresh measurement data arrives. I will consider the job complete when the notebook runs end-to-end on my machine and the validation error stays within a reasonable engineering margin (we can pin down the exact target together at project start).