This project presents the implementation of a Machine Translation system, named ”Implementing Translator from Scratch”, for translating English sentences to Hindi. Leveraging the Transformer architecture, renowned for its attention mechanism, the system achieves efficient handling of long range dependencies. The dataset undergoes comprehensive preprocessing steps, including lowercasing, tokenization, and the addition of start and end tokens. The core of the system comprises Transformer encoder and decoder layers, trained on a dataset of English-Hindi sentence pairs. Extensive training and validation result in accurate translations, as evidenced by low loss values and high BLEU scores. ”Translator from Scratch” highlights the effectiveness of Transformers in Machine Translation tasks and provides insights into future enhancements for such systems.