NVIDIA OpenShell seeks a senior research engineer to identify, prototype, benchmark, and integrate advances in agentic systems and LLM infrastructure into production runtime and developer tools. Requires 8+ years of relevant experience, MS/PhD or equivalent, deep experience with LLMs, agent harnesses, inference infrastructure, evaluation and safety approaches. Base salary range listed: 184,000–287,500 USD/year.

NVIDIA seeks an AI Solutions Architect with 5+ years' experience in AI/software engineering to profile and optimize model training/inference on GPUs, develop and optimize GPU kernels (GEMM, attention), build POCs for LLMs and recommenders, and lead customer-facing technical engagements. Open to remote work with occasional travel; salary range USD 152,000–241,500 per year.

Role involves designing and evaluating routing policies for LLM traffic, building and running agentic benchmarks, contributing code and design docs to open-source, and collaborating with research and engineering teams to integrate optimized inference solutions. Requires 8+ years' experience with deep learning frameworks (PyTorch or TensorFlow), strong ML and systems fundamentals, and experience in model evaluation and benchmarking.

Role to design, implement, optimize, and maintain foundational CUDA C++ libraries, parallel algorithms, utilities, and runtime abstractions. Responsibilities include API and ABI design, performance optimization, testing, profiling, benchmarking, documentation, and collaboration with Python, Rust, compiler, and runtime teams. Minimum qualification: BS/MS/PhD in a related field or equivalent experience and 8+ years of relevant software-development experience; strong production C/C++ skills and practical CUDA or heterogeneous-parallel programming experience.

Role on NVIDIA's High Performance Computing Software development team to develop high-performance communication frameworks and applications. Requires ~5+ years programming in C/C++, 3+ years with Linux, deep knowledge of InfiniBand/Ethernet and computer architecture/operating systems, and experience in performance optimizations. Salary ranges provided for Levels 3 and 4 (USD 152,000–241,500 and USD 184,000–287,500).

Design and implement storage technologies, client libraries, filesystems, and high-performance data paths for GPU-based training and inference; build observability and telemetry for storage bottlenecks and work with platform, SRE, and AI teams to validate systems in production. Requires BS or equivalent and 5+ years of software engineering experience with systems, storage, cloud-native software, and languages such as Go, Python, Rust, C/C++, or Java.

Role involves building systems that gather, aggregate, and normalize topology data from multiple sources, contributing to the open-source Topograph project, and working with NVIDIA GPU hardware to optimize GPU-to-GPU communication and scheduling across cloud providers. Requires strong experience in Go or a systems language, Kubernetes, distributed systems, Slurm/Slinky, Linux, containers, APIs, and CI, and a minimum of 8 years' experience.

Lead performance and scalability analysis and improvements across a Kubernetes accelerated runtime stack (GPU Operator, device plugins, Network Operator, CoCo, etc.), contribute upstream architectural changes, use large-scale simulation infrastructure, and collaborate with researchers, customers, and open-source communities. Requires 5+ years experience, expertise in Kubernetes, distributed accelerator systems, performance optimization, and proficiency in Golang and/or Python.

Own and evolve CI systems that run ML workloads across GPUs and accelerators, improve reliability and reproducibility, apply automation and AI/agent-based workflows to reduce manual operations, and provide a self-service CI platform while partnering with compiler, infrastructure, and release teams. Requires 5+ years' experience and strong Python and CI/platform engineering skills.

Lead technical engagement with CSP/hyperscale customers on GPU firmware and system software integration, drive firmware update orchestration and recovery procedures at rack/ fleet scale, synthesize customer feedback into NVIDIA's firmware/software roadmap, and identify cross-CSP failure patterns to improve tooling, documentation, and tests. Requires 15+ years in GPU system software/firmware or accelerator platform engineering and deep knowledge of GPU internals, multi-GPU fabrics, firmware update lifecycles, error handling, and telemetry.

Lead technical focal point for end-to-end performance in CSP engagements: work with hyperscaler engineering teams to characterize performance, validate benchmarks and tools, gather workload-specific feedback, prioritize optimizations across CUDA/NCCL/driver/firmware, and define test strategies and tooling for NVIDIA certification and customer acceptance. Requires extensive systems performance and GPU/ML infrastructure experience.

Senior technical lead working with CSP engineering teams to ensure rack-scale system software and firmware are deployable and operable at fleet scale. Responsibilities include architecture alignment, driving technical work streams, capturing operational feedback, and improving tooling, monitoring, and update orchestration. Requires 15+ years in system software, firmware, or large-scale distributed systems.

Lead development of scalable mapping and reconstruction systems from multi-camera video and fleet sensor data. Responsibilities include 3D reconstruction, structure-from-motion, pose estimation, map fusion, change detection, sensor fusion, production C++ systems, visualization and evaluation for self-driving applications. Requires 15+ years of experience.

Work on scalable mapping and reconstruction systems for self-driving vehicles, including 3D reconstruction, multi-view geometry, pose estimation, sensor fusion, map fusion, and change detection. Implement production C++ systems and offline pipelines, develop evaluation and visualization tools, and collaborate with perception, localization, simulation, and planning teams. Requires 5+ years experience and strong background in computer vision, geometry, and large-scale sensor data.

Role focuses on leading rack-scale factory and initial data center deployment flows, designing end-to-end factory workflows (firmware flashing, security provisioning, software mitigations), collaborating with data center architects and ODMs/OEMs, and driving pre-silicon readiness using simulation and emulation. Requires 8+ years in system architecture/design and experience with GPUs/DPUs/FPGAs, networking (Ethernet, InfiniBand), and system management protocols (Redfish, IPMI). Salary ranges listed: 184,000–287,500 USD (Level 4) and 224,000–356,500 USD (Level 5).

Role involves co-designing next-generation data center platforms and scalable communications software (e.g., NCCL, NVSHMEM, UCX). Responsibilities include identifying communication bottlenecks, designing and implementing communication technologies, building proofs-of-concept, running experiments, and using simulation and quantitative modeling to evaluate performance at large GPU scales. Requires M.S./Ph.D. or equivalent, 12+ years relevant experience, strong C/C++, parallel programming and communication runtime experience, deep understanding of OS/system/network architecture, and strong Linux skills. Familiarity with CUDA, high-performance networks (InfiniBand, RoCE, NVLink), and deep learning frameworks is a plus.

NVIDIA Cloud Engineering & Services seeks a Principal Software Engineer to mature the APF v0 into a production core platform. Responsibilities include implementing the Runtime Policy Verifier and signed-policy verification, designing deterministic policy projection to runtimes, building conformance and verification suites, collaborating with runtime owners, and driving architecture, versioning, and security specifications. Requires 15+ years experience and strong systems, security, and distributed-systems engineering skills.

NVIDIA seeks a senior software engineer to architect, implement, and optimize production-grade GPU solutions for cuEST, a CUDA-X library for Gaussian-basis quantum chemistry. Requires 10+ years in high-performance numerical methods, 5+ years with production quantum chemistry packages, expert GPU development (C++, CUDA), and a PhD or equivalent experience.

Role requires characterizing and benchmarking large-scale LLM training/inference on GPU/CPU clusters, focusing on networking and communication libraries (NCCL, RDMA, RoCE, MPI). Responsibilities include profiling, developing PyTorch trace-based tools, cross-team performance analysis, and defining test plans. Requires experience with CUDA, deep learning frameworks, containers, and extensive performance evaluation experience.

Work on infrastructure automation across multi-OS, multi-cloud and data center environments using tools such as Ansible, StackStorm, Kubernetes, Docker, Terraform and Jenkins. Requires 3+ years experience, strong scripting skills (Python, Bash, Groovy, Go), CI/CD experience, cloud familiarity (AWS), and experience with agentic AI tools (Codex, Claude).