Fine-tuned GPT-2 model to generate Shakespearean sonnets, demonstrating advanced NLP and transformer architecture expertise through custom training pipeline.

Preprocessed classical poetry dataset and implemented custom tokenization for poetic structure and meter preservation.

Applied transfer learning techniques to adapt pre-trained GPT-2 for creative text generation, optimizing for rhyme scheme and syllable patterns.

Developed comprehensive training pipeline with custom loss functions to maintain sonnet format while ensuring creative output quality.

Developed a robust classification model in Python to identify 153 animal species using ensemble learning, achieving seamless deployment on Hugging Face Spaces.

Enhanced data quality through preprocessing and augmentation techniques, including random flips and crops, to improve model robustness.

Leveraged Transfer Learning with InceptionV3 and Xception models, combining them to boost classification accuracy significantly.

Designed and implemented a convolutional neural network (CNN) from scratch in Python, achieving approximately 85% accuracy in classifying dog and cat images.

Improved model performance by integrating Transfer Learning with pre-trained TensorFlow models, ensuring high accuracy on diverse datasets.

Optimized hyperparameters and applied data augmentation to mitigate overfitting, enhancing the model's generalization capabilities.