NVIDIA has continuously reinvented itself for over two decades. The invention of the GPU in 1999 fueled the growth of PC gaming, redefined modern computer graphics, and revolutionized parallel computing. More recently, GPU deep learning has ignited modern AI, positioning NVIDIA as a leading AI computing company.
There is an increasing focus on delivering AI models locally, closer to the source of data. This reduces latency, improves real-time processing, and addresses privacy concerns by minimizing data transfer to centralized servers. As technology advances, client-side AI (local execution) will play a key role in shaping digital experiences. The LocalAI team is seeking a Systems Software Engineer to build efficient on-device AI software for RTX and DGX-class systems. This role focuses on high-performance local inference, low latency, productive memory use, infrastructure and practical deployment on resource-constrained platforms.
What you'll be doing:
Partnering with NVIDIA's software, research, architecture, and product teams to align strategies and technical needs,
encouraging the ecosystem of AI on RTX and DGX PCs.
Building and optimizing local AI inference stack for RTX, RTX Pro and DGX GPUs, focusing on performance, stability, and scalability across various hardware architectures.
Architecture and development of modern inference runtimes and execution stacks, covering frameworks like Llama.cpp, vLLM, PyTorch, WinML, DXCGC, and TensorRT-RTX across LLMs, vision-language, TTS, ASR, and diffusion AI workloads.
Perform end-to-end optimization of AI models, data pipelines, and inference runtimes to enhance performance across current and next-generation GPU architectures. Apply model optimization techniques such as quantization, pruning, sparsity, and distillation to enable efficient deployment of large models on local and edge devices.
Perform system-level debugging, performance optimization, and performance–accuracy trade-off analysis; develop infrastructure for pe
📌 System Software Engineer (Pune)
🏢 NVIDIA
📍 Pune