A collection of research papers I’ve reproduced, explored, or extended through code.
Implemented a scaled-down Transformer in PyTorch for machine translation. Focused on positional encodings and multi-head attention.
Re-implemented Vision Transformer using PyTorch. The model splits input images into 16x16 patches, encodes them with positional embeddings, and processes them through transformer encoder layers. Evaluated performance on CIFAR-10 and ImageNet subsets.
Reproduced ResNet-18 and ResNet-34 architectures from scratch using PyTorch to understand skip connections and vanishing gradient mitigation. Trained the models on CIFAR-10 and evaluated top-1 accuracy, comparing performance with baseline CNNs.