Predictive Modeling: Optimized a 1-hidden-layer neural network (Adam, LR = 0.15112, 150 epochs) to flag cancer in chest CT scans, hitting 99% train / 98% CV / 99% test accuracy, leaping over past radiologists (~70%) and a logistic regression baseline (76%).

Rigorous Hyper-tuning: Bench-tested optimizers, learning-rates, epochs, and regularizers; Adam beat SGD (Stochastic Gradient Descent) / RMSprop etc , while Learning Rate  0.15112 + 150 epochs lifted CV accuracy from 76% to 96-99% with no over-fitting or under-fitting.

Data Engineering: Curated 899 Kaggle CT images (770 cancer, 129 non-cancer), resized to 64 × 64 RGB, and built stratified 60 / 20 / 20 train-val-test splits to maintain the 6 : 1 imbalance and ensure reliable evaluation.

Healthcare ROI: Early, high-precision detection can slash treatment costs by up to 66%, avert misdiagnosis losses of $90 K – $1.35 M per 1,000 scans, and drive >$700 K annual revenue via faster patient throughput, delivering clear clinical and financial gains.