tpt-infer

Pure-Rust, hardware-agnostic edge AI inference runtime — compile-time tensor shape checking, AOT code generation, zero-allocation execution, and ONNX support. No CUDA, no Python, no C++ toolchains.