Machine Learning Developer

Developed a Neural Machine Translation system focused on translating Malayalam to Dravidian languages using a Transformer-based model. The project involved collecting and preprocessing a large-scale parallel corpus (over 3 crore sentences) using tools like Azure API and Bing Translator. Implemented data cleaning, tokenization with SentencePiece, and vocabulary optimization for more accurate translations. Leveraged the OpenNMT framework and PyTorch to train and fine-tune the model. Evaluated performance using BLEU scores and conducted multi-model comparisons for enhanced accuracy and fluency. The system was internally deployed for testing and demonstrated strong potential for low-resource language translation tasks.