● Advanced Python & OOP Architecture: Mastered Python execution mechanics, closures, functional patterns, encapsulation, polymorphism, abstract classes, and SOLID principles. ● Systems & Data Persistence: Hands-on implementation of file I/O, systems programming concepts, relational database operations with SQL, and SQLite integrations. ● Exploratory Data Analysis & Manipulation: Advanced data wrangling, cleaning, aggregation, and analytical pipelines leveraging NumPy and pandas. ● Classical Machine Learning & Ensembles: Supervised learning (Linear/Logistic Regression, Regularization, SVMs, Decision Trees, Random Forests, Gradient Boosting) and Unsupervised techniques (k-Means, Mixture Models, PCA, Anomaly Detection). ● Deep Learning Foundations & PyTorch: Built custom artificial neural networks (ANN), backpropagation derivations, optimization algorithms, and dynamic computational graphs in PyTorch. ● Computer Vision & Generative AI: CNN architectures, object detection, image segmentation, autoencoders, and diffusion models. ● NLP & Transformer Architecture: Text embeddings, sequential models (RNNs/LSTMs), and self-attention mechanisms in Transformer models.

● Pipeline Engineering: Architected a complete deep learning classification pipeline to detect 5 stages of diabetic retinopathy from fundus imagery. ● Architecture Fine-Tuning: Utilized a pretrained ResNet-50 backbone, replacing the classification head with custom Linear and 50% Dropout layers to enhance generalization. ● Augmentation & Sampling: Addressed class imbalance using stratified sampling, random resized cropping, dynamic rotation, and color space transformations. ● Training Mechanics: Implemented Cross-Entropy Loss with the Adam optimizer (10^-4 learning rate), ReduceLROnPlateau scheduling, and early stopping based on macro F1 score. ● Model Interpretability (XAI): Integrated pytorch_grad_cam to generate visual activation maps on validation samples, confirming model focus on pathology regions.