The combined effect of the plurality of jurisdiction, contexts, and languages has necessitated advanced legal document processing tools to efficiently capture nuances and semantic knowledge (concepts & relations inclusive). This project introduces a novel deep-learning framework designed to address the complexities of legal document translation and summarization, focusing on the Indian legal system and its linguistic diversity. To define a tangible scope, we focus on knowledge representation & corpus curation across all high court judgements in India while targeting abstractive summarization powered by LLMs & translation of English legal documents to Tamil. Our methodology integrates three key innovations: a multi-level Named Entity Recognition (NER) system & knowledge graph generation system for capturing the ontology of legal court documents in the Indian subcontext, a recursive context-aware prompting approach using state-of-the-art LLM for abstractive summarization, and a custom-built Neural Machine Translation (NMT) architecture tailored for English-to-Tamil translation. The NER system employs a hierarchical approach, utilizing Spacy for initial entity detection, a custom Skip-gram model for nuanced term extraction, and advanced LLM-based methods for complex legal relationships. The summarization module leverages state-of-the-art LLMs, optimized through recursive context-aware prompting, to produce contextually rich and condensed summaries.