Recommendation systems are widely used in various online and offline platforms, collaborative filtering being the most commonly used method for implementing them. Most collaborative filtering algorithms, including the ones existing in mlpack, use matrix factorization for this. Although MF is de facto, it cannot deal with complex user-item relationships and also need the user to provide ratings explicitly (explicit feedback). This project proposal deals with these drawbacks by implementing a module for neural collaborative filtering, which can represent complex user-item interaction using neural networks and also can handle implicit feedback (comprising any interaction user has with any item as a positive instance). Neural collaborative filtering, if needed, can also be generalized to matrix factorization. NCF has been proved to give better hit ratio and other metrics than neighborhood based algorithms, and also gives comparable RMSE with existing CF methods in mlpack. This module can give much more flexible recommendations, even when there is less explicit feedback available.