Open-domain Question Answering (QA) systems are meant to answer user questions with an unrestricted amount of knowledge, unlike Closed-domain QA systems that are domain specific and thus have limited knowledge. The existing system requires the user to provide the model with both a question and a paragraph from where the answer can be found; much like asking a model to do reading comprehension. However, a typical user would seek the model for answers to the questions for which the user doesn’t have the knowledge to answer. Hence, we propose a unique methodology for storing the knowledge of the model that allows us to predict answers 50 times faster than the existing system. Then the results are compared with existing models that work on SQuAD datasets and that are meant to handle short contexts (1 or 2 sentences) rather than an entire knowledge base as context. The result indicates that our proposed approach has merit in the open-domain QA system and thus, a better alternative in terms of not just accuracy, but also time complexity.