Quantum devices in the Noisy Intermediate-Scale Quantum (NISQ) era are limited by high error rates and short decoherence times. Typically, compiler optimisations have provided solutions at the gate level. Alternatively, we exploit the finest level of quantum control and introduce a set of pulse level quantum compiler optimisations: sQueeze. Instead of relying on existing calibration that may be inaccurate, we provide a method for the live calibration of two new parameterised basis gates $R_{x}(\theta)$ and $R_{zx}(\theta)$ using an external server. We validate our techniques using the IBM quantum devices and the OpenPulse control interface over more than 8 billion shots. The $R_{x}(\theta)$ gates are on average 52.7% more accurate than their current native Qiskit decompositions, while $R_{zx}(\theta)$ are 22.6% more accurate on average. These more accurate pulses also provide up to a 4.1$\times$ speed-up for single-qubit operations and 3.1$\times$ speed-up for two-qubit gates. Then sQueeze demonstrates up to a 39.6% improvement in the fidelity of quantum benchmark algorithms compared to conventional approaches.