BayLibre Information

BayLibre sits at the crossroads of deep silicon expertise and open source leadership. For more than a decade, we've helped IP vendors, semiconductor designers, module makers, and OEMs turn new hardware into production-ready, adoptable platforms. Our engineering services team is built different: all senior engineers, no leverage model, recruited almost exclusively from the semiconductor industry. Four out of ten hold a community leadership or maintainer role in a FOSS project. What we do: Productizing ML Models & AI Applications — model development, re-training and fine-tuning for Cloud & Edge; quantizing and right-sizing for edge AI hardware; application architecture spanning Cloud and Local AI within SWaP-TC budgets (size, weight, power, thermals, cost); ROS application architecture & development. Compilers & ML Toolchains — compiler backends and codegen for RISC-V, NPU/VPU/APU and datacenter coprocessors; GPU offload for HPC via OpenMP/OpenACC; toolchain enablement for PyTorch, ExecuTorch and LiteRT; custom IRs and kernels linking the ML framework value chain. Operating Systems & System Integration — firmware for novel AI accelerator IP; ROS middleware integration & acceleration; upstream-first, production-grade bootloaders, kernels, DeviceTree, drivers and middleware; secure OTA, CRA compliance, virtualization and sustaining engineering. Moonshots / R&D — tackling the hardest engineering challenges from our most demanding customers, working at the specification level, and innovating on product form factors that are new entrants to the market. All of the above has been upstreamed by our team — one of the strengths of our reputation in the FOSS community, and a reflection of the leadership roles we hold in key software projects. You'll find us leading the charge in PyTorch & ExecuTorch; LLVM, MLIR & IREE; GCC & GOMP (OpenMP & OpenACC); ROS2; the Linux kernel; Zephyr RTOS; the Yocto Project; and AOSP. Together, we're building something great.