TranslingoX: Real-Time Translation System for Indian Languages

India's linguistic diversity creates communication challenges throughout the multilingual population. Traditional translation systems often do not perform real-time translation, have little contextual reasoning, or are limited in their support for low-resource Indian languages. This work proposes a real-time translation architecture specially developed for Indian languages using a hybrid deep learning framework combining Transformer-based Neural Machine Translation (NMT) and contextual embeddings via multilingual BERT. This system enables crossword translations among the most common Indian languages: Hindi, Tamil, Telugu, Bengali. The NMT model solves basic grammatical structure and translation fluency, whereas contextual embeddings build a strong semantic representation to handle dialectal variations. Also integrated are speech properties of the system that convert speech to text and text to speech so that users can interact with each other via voice communication in real time. The solution becomes IoT-enabled for mobile and edge device deployments, while usage will be targeted toward education, health, and governance. This module achieves good experimental results exhibiting high BLEU scores and low latency and proves better efficiency in live scenarios.