The field of Natural Language Processing (NLP) is moving towards increased support for low-resource languages, with a focus on developing scalable and efficient models that can be easily adapted across languages. Recent work has highlighted the importance of modular and open-source design, allowing for rapid cross-lingual adaptation and reducing model design and tuning overhead. The use of transformer-based architectures has shown promising results, with the ability to achieve high accuracy across multiple languages and tasks. Additionally, techniques such as automatic mixed precision training have been explored to improve computational efficiency without sacrificing model performance. Notable papers include: FastPOS, which proposes a language-agnostic POS tagging framework that achieves high accuracy in low-resource languages. Accelerating Bangla NLP Tasks with Automatic Mixed Precision, which demonstrates the potential of automatic mixed precision training to accelerate training and reduce memory consumption in NLP tasks. Feature Selection Empowered BERT, which presents a data-efficient strategy for fine-tuning BERT on hate speech classification tasks. Bangla Hate Speech Classification with Fine-tuned Transformer Models, which studies the use of transformer-based models for hate speech classification in the low-resource Bangla language.