The paper introduces AlphaTensor-Quantum, a method for optimizing the number of T gates in quantum circuits, a critical aspect of fault-tolerant quantum computing. AlphaTensor-Quantum leverages deep reinforcement learning and tensor decomposition to minimize T-count, a key metric for circuit complexity. Unlike existing methods, it incorporates domain-specific knowledge and *gadgets*—constructs that reduce the number of T gates—significantly improving efficiency. The method outperforms existing T-count optimization techniques on arithmetic benchmarks and finds efficient algorithms for multiplication in finite fields, matching the complexity of classical Karatsuba's method. It also optimizes circuits for Shor's algorithm and quantum chemistry simulations, demonstrating its ability to automate the design of optimized circuits. AlphaTensor-Quantum's flexibility allows for future extensions to optimize other metrics and handle different types of gates, making it a promising tool for advancing quantum computing research and applications.The paper introduces AlphaTensor-Quantum, a method for optimizing the number of T gates in quantum circuits, a critical aspect of fault-tolerant quantum computing. AlphaTensor-Quantum leverages deep reinforcement learning and tensor decomposition to minimize T-count, a key metric for circuit complexity. Unlike existing methods, it incorporates domain-specific knowledge and *gadgets*—constructs that reduce the number of T gates—significantly improving efficiency. The method outperforms existing T-count optimization techniques on arithmetic benchmarks and finds efficient algorithms for multiplication in finite fields, matching the complexity of classical Karatsuba's method. It also optimizes circuits for Shor's algorithm and quantum chemistry simulations, demonstrating its ability to automate the design of optimized circuits. AlphaTensor-Quantum's flexibility allows for future extensions to optimize other metrics and handle different types of gates, making it a promising tool for advancing quantum computing research and applications.