Build faster. Deploy smarter.
One SDK to build, debug, and deploy across CPUs, GPUs, FPGAs, and Kalray DPUs—flexibly routing each stage to the right processor for higher perf/watt and predictable workflows from dev to production.
The multi-accelerator challenge
Traditional workflows force teams to juggle different models and toolchains for each device. The result: steep learning curves, fragmented code, wasted cycles, and fragile deployments. Brane SDK removes that friction with a flexible, unified way to build once and run across CPUs, GPUs, FPGAs, and Kalray DPUs.
One environment
Build, debug, and package in a single IDE—no juggling ROCm/CUDA/FPGA toolchains.
Device-aware execution
Route each stage to the best processor (CPU, GPU, DPU, FPGA) to maximize perf per watt.
Reproducible deploys
Containerized runners, vetted drivers/firmware, versioned configs—dev → edge without surprises.
One SDK, multiple accelerators. The Brane SDK eliminates the complexity of managing separate toolchains for CPUs, GPUs, FPGAs, and DPUs. Build once and deploy across your entire heterogeneous computing infrastructure with a single, unified development environment.
One Development Environment
Build with familiar tools – Gradle-based compilation works with any SDK, while IntelliJ integration provides AI plugins and advanced debugging.
Cross-Platform Compatibility
Write once, deploy across multiple hardware accelerators with consistent APIs and predictable performance.
AI-Powered Development
IntelliJ AI plugins provide code optimization suggestions, performance analysis, and intelligent debugging assistance.
Supported Hardware
- GPU: AMD Instinct (ROCm), NVIDIA (CUDA on request)
- DPU: Kalray TC4 for predictable parallel processing
- FPGA: AMD/Xilinx AI FPGAs for custom acceleration
- CPU: x86-64 orchestration and general computing
The Brane SDK delivers measurable advantages across development velocity, system performance, and operational reliability—enabling teams to focus on innovation rather than integration complexity.
Development Velocity
Hardware Flexibility
Operational Control
Key Outcomes
AccelOne SDK & AI Services
Start with the AccelOne SDK for unified multi-accelerator development, then add on-demand AI agents or custom operators when you’re ready to scale.
AccelOne SDK — Developer Toolkit
Developer Access
Unified development environment that orchestrates CPU, GPU, DPU, and FPGA resources from a single toolchain. Build once, deploy across your entire multi-accelerator infrastructure with containerized reproducibility.
Key Features
- One IDE/CLI for build, debug, deploy across all accelerators
- Target CPU, GPU (ROCm/CUDA), DPU (Kalray), and FPGA
- Containerized toolchains for reproducible builds
- Pipelines & operators with device-aware routing
AI Agents — On-Demand
Services
Custom AI agents designed for your specific workflows, packaged with AccelOne SDK pipelines and observability. From RAG systems to orchestration agents, deployed on-premises with enterprise security.
What We Deliver
- Task-specific agents (RAG, extraction, eval, orchestration)
- Packaged with AccelOne SDK pipelines & observability
- On-prem or cloud deploys with security-first design
- Custom operators & hardware offload (GPU/DPU/FPGA)
Want to compare offerings or bundle with hardware? Explore our full catalog.
Visit the ShopBrane SDK — FAQ
Quick answers on capabilities, platforms, and how to get started with multi-accelerator development.
What is the Brane SDK?
A unified development kit for building applications and software that run across CPUs, GPUs, DPUs, and FPGAs. Based on Gradle and IntelliJ technology with AI-powered development assistance to accelerate your heterogeneous computing projects.
Which accelerators and stacks are supported?
CPU: x86_64 architecture. GPU: AMD ROCm by default; CUDA on request. DPU: Kalray TC4. FPGA: AMD/Xilinx and Intel FPGAs. You can mix devices in one application; the runtime coordinates execution across accelerators.
What OS and tools do I need?
Linux (Ubuntu LTS recommended) for full development capabilities. Windows 11 Pro supported for CPU/GPU workflows; Kalray TC4 requires Linux. The SDK integrates with familiar tools like Gradle and IntelliJ for seamless development.
Can I integrate my own operators or third-party libraries?
Yes. Implement custom operators against the SDK’s unified API and target different accelerators (CPU/GPU/DPU/FPGA). Common ML libraries (e.g., PyTorch/ONNX/TensorRT) can be integrated within your applications where applicable.
How do I get access?
Request Developer Access and tell us about your use case, target hardware, and development needs. We’ll provide SDK access and starter resources. Request Dev Access or Talk to an Engineer.
Ready to start building with the AccelOne SDK?
Get hands-on support for multi-accelerator development. Whether you’re building new applications or porting existing code to run across CPUs, GPUs, DPUs, and FPGAs, our team will help you leverage the full power of heterogeneous computing.
- SDK setup and IntelliJ integration with AI plugins
- Cross-platform development best practices
- Hardware configuration guidance (3x GPU, GPU+DPU, GPU+FPGA)
- Performance optimization and troubleshooting