NVIDIA is hiring a Senior Deep Learning Inference Software Engineer to design, build, and optimize GPU-accelerated software and inference frameworks (vLLM, SGLang, FlashInfer). The role centers on performance tuning for LLMs, multimodal and generative AI across NVIDIA accelerators and requires strong C/C++ skills, 5+ years' software development experience, and GPU/CUDA experience. Salary ranges for Poland are provided.

You will author merged PRs and build integrations (libraries, agent frameworks, MCP servers, automation platform connectors) so developers and AI agents can access SearchApi. Required: proven merged open-source contributions in repos you don't own, fluency with Ruby/Python/TypeScript and at least one agent/automation framework, and strong written English for docs and PRs.

Role to optimize inference speed and carry out production deployment of Cosmos foundation models on diverse GPU platforms; profile and remove bottlenecks in deep learning workloads. Requires 5+ years' experience, MSc/PhD or equivalent, strong deep learning background, Python and PyTorch skills, and experience with inference optimization techniques and frameworks (e.g., TensorRT, vLLM).