Vaccinations against viruses have always been the need of the hour since long past. However, it is hard to efficiently distribute the vaccines (on time) to all the corners of a country, especially during a pandemic. Considering the vastness of the population, diversified communities, and demands of a smart society, it is an important task to optimize the vaccine distribution strategy in any country/state effectively. Although there is a profusion of data (Big Data) from various vaccine administration sites that can be mined to gain valuable insights about mass vaccination drives, very few attempts has been made towards revolutionizing the traditional mass vaccination campaigns to mitigate the socio-economic crises of pandemic afflicted countries. In this paper, we bridge this gap in studies and experimentation. We collect daily vaccination data which is publicly available and carefully analyze it to generate meaning-full insights and predictions. We put forward a novel framework leveraging Supervised Learning and Reinforcement Learning (RL) which we call VacciNet, that is capable of learning to predict the demand of vaccination in a state of a country as well as suggest optimal vaccine allocation in the state for minimum cost of procurement and supply. At the present, our framework is trained and tested with vaccination data of the USA.