Zml

ZML is a Paris-based startup, founded in 2023 by Steeve Morin, that builds a production inference stack decoupling AI workloads from proprietary hardware. Its 'Model to Metal' approach compiles models directly to NVIDIA, AMD, TPU, and Trainium accelerators from a single codebase, avoiding Python runtime overhead and hidden abstractions to achieve peak hardware performance. The company recently released ZML/LLMD, a tool that lets developers run open-source AI models across different chip types. ZML has raised a seed round backed by investors including 20VC Fund and Kindred Capital.