This study presents a novel approach for translating Maharashtri Prakrit (Mahārāṣṭrī Prākṛta), an ancient Indo-Aryan language, to English using neural machine translation. This paper discusses the challenges faced in creating datasets for Maharashtri Prakrit, due to it being an extremely low-resource language lacking digitized resources. Also, it covers many neural machine translation methods which are discussed in this paper. Out of these methods, Maharashtri Prakrit was translated by fine-tuning Facebook’s M2M100 model, a multilingual model that can translate over 100 languages. This model was fine-tuned to the dataset created by collecting Maharashtri Prakrit and its corresponding translations from various sources. The model’s performance was evaluated using BLEU and METEOR scores to evaluate the accuracy and fluency of translations, which resulted in a BLEU score of 15.3416 and a METEOR score of 0.4723. By providing a model for translating Maharashtri Prakrit, this paper aims to preserve India’s heritage and make this language more accessible to the general public, historians, and linguists.