Scope & Responsibilities

• Explored core Transformer models, language modeling objectives, and generative text architectures.
• Prepared custom textual datasets, engineered tokenization pipelines, and managed tensor format dimensions.
• Fine-tuned pre-trained model weights (GPT-2) locally using Hugging Face Transformers, tracking training loss and convergence metrics.
• Evaluated model checkpoints using prompt completion inference to assess coherence and contextual similarity to training distributions.
• Documented optimization parameters, learning rate schedules, and temperature parameters for downstream text generation.