Real-time personalization engines enable effective customization in e-commerce. Yet, the development of such engines is not trivial. It remains challenging to optimize across many options, especially while utilizing context information in real time. To meet these challenges, we aim to provide an easy-to-implement personalization engine to support adaptive treatment assignment in sequential experiments. We formalize the personalization problem under the multi-armed bandit framework, and propose a new contextual bandit algorithm based on the particle filtering technique. Our method allows firms to flexibly introduce new personalized options, calibrate their impact using prior knowledge from historical data and rapidly update these prior beliefs as new observations arrive. With an application to news-article recommendation, we show that the proposed method achieves a Click-Through-Rate (CTR) of 5.23%, outperforming state-of-the-art methods like UCB and Linear-regression-based UCB that achieve a CTR of 4.53% and 4.82%, respectively.