Nvidia Senior On-Device Model Inference Optimization Engineer 

China, Beijing, Beijing 

919717667

What you'll be doing:

• Develop and implement strategies to optimize AI model inference for on-device deployment.

• Employ techniques like pruning, quantization, and knowledge distillation to minimize model size and computational demands.

• Optimize performance-critical components using CUDA and C++.

• Collaborate with multi-functional teams to align optimization efforts with hardware capabilities and deployment needs.

• Benchmark inference performance, identify bottlenecks, and implement solutions.

• Research and apply innovative methods for inference optimization.

• Adapt models for diverse hardware platforms and operating systems with varying capabilities.

• Create tools to validate the accuracy and latency of deployed models at scale with minimal friction.

• Recommend and implement model architecture changes to improve the accuracy-latency balance.

What we need to see:

• MSc or PhD in Computer Science, Engineering, or a related field, or equivalent experience.

• Over 5 years of confirmed experience specializing in model inference and optimization.

• 8+ overall years of work experience in a relevant area

• Expertise in modern machine learning frameworks, particularly PyTorch, ONNX, and TensorRT.

• Proven experience in optimizing inference for transformer and convolutional architectures.

• Strong programming proficiency in CUDA, Python, and C++.

• In-depth knowledge of optimization techniques, including quantization, pruning, distillation, and hardware-aware neural architecture search.

• Skilled in building and deploying scalable, cloud-based inference systems.

• Passionate about developing efficient, production-ready solutions with a strong focus on code quality and performance.

• Meticulous attention to detail, ensuring precision and reliability in safety-critical systems.

• Strong collaboration and communication skills for working optimally across multidisciplinary teams.

• A proactive, diligent mentality with a drive to tackle complex optimization challenges.

Ways to stand out from the crowd:

• Publications or industry experience in optimizing and deploying model inference at scale.

• Hands-on expertise in hardware-aware optimizations and accelerators such as GPUs, TPUs, or custom ASICs.

• Active contributions to open-source projects focused on inference optimization or machine learning frameworks.

• Experience in designing and deploying inference pipelines for real-time or autonomous systems.