Deep Learning Performance Architect jobs at Nvidia in China, Shanghai

NVIDIA has been transforming computer graphics, PC gaming, and accelerated computing for more than 25 years. It’s a unique legacy of innovation that’s fueled by great technology—and amazing people.

• Design and implement the DSL and the core compiler of tile-aware GPU programming model for emerging GPU architectures

• Continuously innovate and iterate on the core architecture of the compiler to consistently optimize performance

• Investigation of next-generation GPU architectures and provide solutions in the DSL and compiler stack

• Performance analysis on emerging AI/LLM workloads and integrate with AI/ML frameworks

• Masters or PhD or equivalent experience in relevant discipline (CE, CS&E, CS, AI)

• 4 + years of relevant work experience

• Excellent C/C++ programming and software engineering skills, ACM background is a plus

• Good fundamental knowledges on computer architecture

• Strong ability in abstracting problems and the methodology in resolving problems

• Strong compiler backgrounds including MLIR/TVM/Triton/LLVM is desired

• Good knowledge of GPU architecture and fast kernel programming skills is a plus

• Knowledge of LLM algorithms or a certain HPC domain is a plus

• Knowledge of multi-GPU distributed communication is a plus

• Excellent oral communication in English is a plus

What you'll be doing:

• Developing and introducing groundbreaking reinforcement learning algorithms tailored for LLM applications.

• Collaborating with a world-class team of engineers and researchers to integrate these algorithms into applied scenarios.

• Using your extensive expertise in math and AI to improve the reasoning capabilities of our models.

• Engaging in rigorous testing and refinement processes to ensure flawless performance and reliability.

• Contributing to our collective goal of delivering industry-leading AI solutions, strictly adhering to NVIDIA's high standards.

What we need to see:

• Proficient in C++/Python programming.

• 3+ years working experience.

• BS or MS (or equivalent experience) in CS, CE, EE, or a related field.

• Proven experience in reinforcement learning and its application to large language models.

• Strong background in mathematics and AI algorithms, with a focus on reinforcement learning.

• Demonstrated history of applying reinforcement learning algorithms in practical scenarios.

• Understanding of GPU architecture is a huge plus.

• Excellent problem-solving skills and the ability to work collaboratively in a dynamic team environment.

• A passion for innovation and a dedication to achieving outstanding results.

What you will be doing:

• Develop and optimize the control stack, including locomotion, manipulation, and whole-body control algorithms;

• Deploy and evaluate neural network models in physics simulation and on real humanoid hardware;

• Design and maintain teleoperation software for controlling humanoid robots with low latency and high precision;

• Implement tools and processes for regular robot maintenance, diagnostics, and troubleshooting to ensure system reliability;

• Monitor teleoperators at the lab and develop quality assurance workflows to ensure high-quality data collection;

• Collaborate with researchers on model training, data processing, and MLOps lifecycle.

What we need to see:

• Bachelor’s degree in Computer Science, Robotics, Engineering, or a related field;

• 3+ years of full-time industry experience in robotics hardware or software full-stack;

• Hands-on experience with deploying and debugging neural network models on robotic hardware;

• Ability to implement real-time control algorithms, teleoperation stack, and sensor fusion;

• Proficiency in languages such as Python, C++, and experience with robotics frames (ROS) and physics simulation (Gazebo, Mujoco, Isaac, etc.).

• Experience in maintaining and troubleshooting robotic systems, including mechanical, electrical, and software components.

• Physically work on-site on all business days.

Ways to stand out from the crowd:

• Master’s or PhD’s degree in Computer Science, Robotics, Engineering, or a related field;

• Experience at humanoid robotics companies on real hardware deployment;

• Experience in robot hardware design;

• Demonstrated Tech Lead experience, coordinating a team of robotics engineers and driving projects from conception to deployment.

What you'll be doing:
• Work on research, design and implementation of software features that are beneficial to our customers to meet their performance targets and build unique values

• World model and nvidia cosmos related development and bug fixes

• Develop solutions for DNN models acceleration optimization and deployment.

What we need to see:
• Subject to arrangement to different works at any time to take on different tasks and challenges

• Self-motivated attitude and motivation to make things success, intent to learn

• BS/MS degree in Computer Science/EE or related

• Proven fundamentals in c++/python programming and SW design and debug skills

• Abundant knowledge of ML/DL techniques for Computer Vision and autonomous driving

• Believe that experimenting is the only way to find the truth, not argument and arbitary guess

• Located in Shanghai or willing to work in Shanghai

Ways to stand out from the crowd:
• DNN development and network acceleration is highly desired

• Familiarity with GPU computing/NVIDIA CUDA/NVIDIA TensorRT

What you will be doing:

• Collaborate closely with our AI/ML researchers to make their ML models more efficient leading to significant productivity improvements and cost savings

• Build tools, frameworks, and apply ML techniques to detect & analyze efficiency bottlenecks and deliver productivity improvements for our researchers

• Work with researchers working on a variety of innovative ML workloads across Robotics, Autonomous vehicles, LLM’s, Videos and more

• Collaborate across the engineering organizations to deliver efficiency in our usage of hardware, software, and infrastructure

• Proactively monitor fleet wide utilization patterns, analyze existing inefficiency patterns, or discover new patterns, and deliver scalable solutions to solve them

• Keep up to date with the most recent developments in AI/ML technologies, frameworks, and successful strategies, and advocate for their integration within the organization.

What we need to see:

• BS or similar background in Computer Science or related area (or equivalent experience)

• Minimum 8+ years of experience designing and operating large scale compute infrastructure

• Strong understanding of modern ML techniques and tools

• Experience investigating, and resolving, training & inference performance end to end

• Debugging and optimization experience with NSight Systems and NSight Compute

• Experience with debugging large-scale distributed training using NCCL

• Proficiency in programming & scripting languages such as Python, Go, Bash, as well as familiarity with cloud computing platforms (e.g., AWS, GCP, Azure) in addition to experience with parallel computing frameworks and paradigms.

• Dedication to ongoing learning and staying updated on new technologies and innovative methods in the AI/ML infrastructure sector.

• Excellent communication and collaboration skills, with the ability to work effectively with teams and individuals of different backgrounds

Ways to stand out from the crowd:

• Background with NVIDIA GPUs, CUDA Programming, NCCL and MLPerf benchmarking

• Experience with Machine Learning and Deep Learning concepts, algorithms and models

• Familiarity with InfiniBand with IBOP and RDMA

• Understanding of fast, distributed storage systems like Lustre and GPFS for AI/HPC workloads

• Familiarity with deep learning frameworks like PyTorch and TensorFlow

What you'll be doing:

• Developing and introducing groundbreaking reinforcement learning algorithms tailored for LLM applications.

• Collaborating with a world-class team of engineers and researchers to integrate these algorithms into applied scenarios.

• Using your extensive expertise in math and AI to improve the reasoning capabilities of our models.

• Engaging in rigorous testing and refinement processes to ensure flawless performance and reliability.

• Contributing to our collective goal of delivering industry-leading AI solutions, strictly adhering to NVIDIA's high standards.

What we need to see:

• Masters or PhD or equivalent experience in relevant discipline (CE, CS&E, CS, AI)

• Proficient in C++/Python programming.

• Proven experience in reinforcement learning and its application to large language models.

• Strong background in mathematics and AI algorithms, with a focus on reinforcement learning.

• Demonstrated history of applying reinforcement learning algorithms in practical scenarios.

• Understanding of GPU architecture is a huge plus.

• Excellent problem-solving skills and the ability to work collaboratively in a dynamic team environment.

• A passion for innovation and a dedication to achieving outstanding results.

What you’ll be doing:

• Analyze state of the art DL networks (LLM etc.), identify and prototype performance opportunities to influence SW and Architecture team for NVIDIA's current and next gen inference products.

• Develop analytical models for the state of the art deep learning networks and algorithm to innovate processor and system architectures design for performance and efficiency.

• Specify hardware/software configurations and metrics to analyze performance, power, and accuracy in existing and future uni-processor and multiprocessor configurations.

• Collaborate across the company to guide the direction of next-gen deep learning HW/SW by working with architecture, software, and product teams.

What we need to see:

• BS or higher degree in a relevant technical field (CS, EE, CE, Math, etc.).

• Strong programming skills in Python, C, C++.

• Strong background in computer architecture.

• Experience with performance modeling, architecture simulation, profiling, and analysis.

• Prior experience with LLM or generative AI algorithms.

Ways to stand out from the crowd:

• GPU Computing and parallel programming models such as CUDA and OpenCL.

• Architecture of or workload analysis on other deep learning accelerators.

• Deep neural network training, inference and optimization in leading frameworks (e.g. Pytorch, TensorRT-LLM, vLLM, etc.).

• Open-sourceAIcompilers (OpenAI Triton, MLIR, TVM, XLA, etc.).