In this era of noisy intermediate scale quantum devices, variational quantum algorithms show promise as an avenue for quantum advantage. The performance of such approaches for specific quantum applications is closely coupled to the details of the quantum hardware with respect to the noise levels and the variational quantum circuit. To achieve optimal performance, the choice of variational quantum circuit architecture should therefore be informed by both the hardware details as well as the goal of the application. In other words, co-design of quantum applications for their specific hardware and at all levels of the quantum computing stack should be explored to achieve optimal performance. In this work, we propose a reinforcement learning algorithm that co-designs noise resilient quantum circuits and qubit mappings based on the knowledge of chemistry problems, quantum architectures and noise models. To reduce misalignment between chemistry problems' resource requirements and existing quantum hardware and enable simulation of larger chemical systems on near term quantum computers. We demonstrate the performance of our algorithm on the electronic structure problem of the water molecule.