Identifying the required skill set for a given job is a crucial task for both the job market and individual candidates. It has recently become increasingly complex due to both the widening of the set of available skills as well as the diverse and confusing language within the job postings. This paper aims to help people looking for a job by suggesting the set of skills that they would need for their next move in the job market. To this end, we propose a Convolutional-based Job Skill Predictor (CJSP) that utilizes convolutional neural networks and pretrained word vectors to determine which skills a given person should pursue. We study how different embedding matrices work in combination with textual features and how they affect the performance of our system. We also propose a Dense-based Job Skill Predictor (DJSP) that is an adapted version of CJSP without convolutional layers, which incorporates taxonomy-based features derived from WORDNET and custom ESCO data. The experimental results show great potential for improving the job hunting workflow of both individuals and HR, by streamlining job matchmaking and assisting workers in skill-based job transitions.