Senior System Software Engineer - LocalAI

About the role

NVIDIA has continuously reinvented itself for more than two decades. The invention of the GPU in 1999 fueled the growth of the PC gaming market, redefined modern computer graphics, and revolutionized parallel computing. GPU-powered deep learning helped ignite the era of modern AI, establishing GPUs as the foundation of intelligent applications across productivity, gaming, and creative workflows, and reinforcing NVIDIA’s position as a leading AI computing company.

More recently, there is a growing focus on running AI models locally, closer to where data is generated. This approach reduces latency, enables real-time processing, and addresses privacy concerns by minimizing the need to send data to centralized servers. As technology continues to evolve, client-side AI will play an increasingly important role in shaping the digital landscape. The LocalAI team is seeking a Senior Systems Software Engineer to develop efficient on-device AI software for RTX and DGX-class systems. The role focuses on delivering high-performance local inference with low latency, optimized memory utilization, robust infrastructure, and practical deployment on resource-constrained platforms.

What You’ll Be Doing

• Partner with NVIDIA’s software, research, architecture, and product teams to align technical requirements and strategic priorities, fostering the AI ecosystem on RTX and DGX PCs.

• Build and optimize the local AI inference stack for RTX, RTX Pro, and DGX GPUs, with a focus on performance, stability, and scalability across diverse hardware architectures.

• Design and develop modern inference runtimes and execution stacks using frameworks such as llama.cpp, vLLM, PyTorch, WinML, DXCGC, and TensorRT-RTX, supporting LLM, vision-language, TTS, ASR, and diffusion-based AI workloads.

• Perform end-to-end optimization of AI models, data pipelines, and inference runtimes to maximize performance on current and next-generation GPU architectures. Apply model optimization techniques, including quantization, pruning, sparsity, and distillation, to enable efficient deployment of large models on local and edge devices.

• Conduct system-level debugging, performance tuning, and performance-accuracy trade-off analysis; develop infrastructure for performance and accuracy sweeps; analyse results to identify gaps and drive fixes; and establish engineering guidelines to accelerate bring-up and ensure production readiness of new models and inference backends.

What we need to see

• 5+ Years of experience with Bachelor’s, Master’s, or PhD in Computer Science, Software Engineering, Mathematics, or a related field, or equivalent experience.

• Excellent C++ programming and debugging skills, with a strong foundation in data structures, algorithms, and machine learning.

• Proven experience developing and optimizing AI inference pipelines and applications using ML/DL frameworks such as Llama.cpp, vLLM, PyTorch, Windows ML, DXCGC, and TensorRT.

• Deep understanding of inference backends and runtime internals, including scheduling, memory management, KV-cache behaviour, graph execution, quantization, and hardware-aware optimization.

• Strong analytical and problem-solving skills, with the ability to manage multiple priorities effectively in a fast-paced environment.

• Excellent written and verbal communication skills, enabling effective collaboration across engineering teams and management.

Ways to stand out from the crowd

• Understanding of modern machine learning, deep neural network, and generative AI techniques, with relevant contributions to major open-source projects.

• Consistent track record of delivering end-to-end products in multinational companies with geographically distributed teams.

• Proficiency in low-level system and GPU programming, CUDA, and the development of high-performance systems.

• Contributions to open-source inference runtimes, model tooling, or performance infrastructure.

• Hands-on experience building applications using frameworks and APIs such as llama.cpp, PyTorch, TensorRT, Vulkan, DirectX, and vLLM.