Recent research indicates that large language models lack consistent reliability in tasks requiring complex reasoning. While they may impress us with fluently written articles prompted by user input, they can easily disappoint by displaying shortcomings in basic reasoning skills, such as the functional understanding of ‘left is the opposite of right’, let alone grounding such concepts in diverse real-world situations involving perception and action. To address real-world problems, computational models often need to involve multiple interdependent learners, along with significant levels of composition and reasoning. In this talk, I will present our findings regarding reasoning challenges of LLMs and discuss how symbolic representations can leverage the capacity of neural models for compositional reasoning over complex linguistic structures, grounding language in visual perception, combining multiple modalities of information and handling uncertainty. I will highlight our efforts in Neurosymbolic modeling and introduce DomiKnowS, our developed library that facilitates such modeling. The DomiKnowS framework exploits both symbolic and sub-symbolic representations to solve complex, AI-complete problems and seamlessly integrates symbolic and logical knowledge into deep models through various underlying algorithms.