GPU and edge AI systems

Compute and software choices assessed against useful model behaviour and the complete serving workload.

Enterprise engineering field

Compute and software choices assessed against useful model behaviour and the complete serving workload.

The engineering field

Compute and software choices assessed against useful model behaviour and the complete serving workload. Softinator connects the selected platform to the surrounding application, information and operating requirements.

Choose the platform

This field includes NVIDIA AI Enterprise, NVIDIA NIM, NVIDIA TensorRT, NVIDIA TensorRT LLM, NVIDIA Triton Inference Server, NVIDIA NeMo, NVIDIA NeMo Guardrails. Each platform page connects its particular engineering context to relevant commissioned services. Use the platform links below to identify the ecosystem suited to the requirement.

Define the engagement

The programme establishes components, interfaces, responsibilities and acceptance evidence. Vendor access and licensing follow the selected scope. Relevant project material remains tied to its own contribution and permitted publication context.

Platforms in this field.

AMD ROCm engineeringGPU compute and model workload integration. Selected hardware compatibility and workload assessment.AMD Versal engineeringAdaptive compute and embedded application integration. Hardware-aware task processing and deployment boundaries.Hailo AI processors engineeringEdge inference hardware integration. Model compatibility and device-constrained visual tasks.Intel OpenVINO engineeringModel inference on selected Intel deployment targets. Model conversion and task-specific device behaviour.NVIDIA AI Enterprise engineeringEnterprise software for AI workload deployment. Selected software components, support scope and operating requirements.NVIDIA BlueField engineeringData-centre processing and infrastructure interfaces. Workload networking, isolation and operating requirements.NVIDIA CUDA engineeringGPU application and compute development. Workload kernels and device-specific execution constraints.NVIDIA DeepStream engineeringVideo and vision inference pipelines. Stream observations and edge task integration.NVIDIA DGX engineeringIntegrated GPU infrastructure for AI programmes. Training and serving workloads with complete operating dependencies.NVIDIA Jetson Orin engineeringEdge compute for robot and vision applications. Device resources, model packages and sensing interfaces.NVIDIA Jetson Thor engineeringEdge compute for advanced robot workloads. Selected embodied models and device-level task requirements.NVIDIA NeMo Guardrails engineeringControlled conversational and agent behaviour. Task boundaries and policy-relevant application responses.NVIDIA NeMo engineeringModel development and adaptation workflows. Task-specific training contributions and evaluation boundaries.NVIDIA NIM engineeringPackaged model inference workflows. Model endpoints and enterprise application integration.NVIDIA RAPIDS engineeringGPU-accelerated data and analytical workloads. Selected data-processing tasks and infrastructure fit.NVIDIA Spectrum engineeringData-centre networking for distributed workloads. Network capacity and dependencies around GPU execution.NVIDIA TensorRT LLM engineeringLanguage-model inference optimisation. Serving workload and quality-versus-capacity tradeoffs.NVIDIA TensorRT engineeringOptimised inference engine development. Model-to-engine conversion and hardware-specific task quality.NVIDIA Triton Inference Server engineeringModel-serving interfaces and workload operation. Endpoint integration and multi-model serving behaviour.Qualcomm AI Hub engineeringDevice-focused model optimisation and deployment workflows. Target-device model preparation and task validation.

Bring the next
requirement.

Discuss your programme