Book excerpt: Quantum computing is a type of computation that harnesses laws of quantum mechanics such as superposition and entanglement to solve problems that are too complex for classical computers. Theoretically, for instance, Shor's algorithm brings about almost exponential acceleration for finding the prime factorization of an integer compared to the most efficient known classical algorithm. However, such quantum computational advantage is largely restricted by near-term quantum computers which provide only a limited number of qubits and suffer from various types of noises, such as decoherence, gate errors, measurement errors, and crosstalk, etc. Quantum computing advantage is currently mostly demonstrated on specifically designed sampling tasks, thereby making little societal impact through practical applications. Despite quantum hardware limitations, hybrid quantum-classical algorithms have recently been proposed to exploit possible quantum computation advantages in multiple fields such as quantum machine learning and optimization which are less impacted by quantum noises. Classical machine learning and optimization have been transforming many walks of our lives, from intelligent transportation, to automated industrial decision making and operation, to AI-driven drug discovery and development. Quantum machine learning and optimization could leverage the mentioned quantum phenomena and empower some classical algorithms to compute more efficiently or achieve better performance. Hybrid algorithms are promising approaches to combine the computational advantages from quantum and classical machines in practical applications. I studied quantum machine leaning and optimization approaches for utilizing quantum computational advantages in societal applications, especially in autonomous driving and drug discovery during my Ph.D. More specifically, quantum approximate optimization was investigated on quantum machines with different qubit technologies for object detection applications. Multiple quantum generative models were developed and examined for drug discovery. Apart from these quantum machine learning approaches, a scalable quantum optimization algorithm was designed with divide-and-conquer paradigm for solving some large-scale combinatorial optimization tasks even on near-term quantum computers.