VoltaML is a lightweight library that accelerates machine learning and deep learning model inference through optimization and compilation. It supports multiple inference runtimes including TensorRT, TorchScript, ONNX, and TVM, and provides out-of-the-box support for FP16 and Int8 quantization. The library offers hardware-specific compilation targeting CPU and GPU devices with deployment achievable through a single line of code.