Senior Software Engineer, Deep Learning Inference

NVIDIA has been at the forefront of the deep learning revolution, pioneering innovations that have transformed the entire field. As the leading provider of GPUs and AI computing platforms, NVIDIA has empowered researchers and engineers worldwide to accelerate breakthroughs in artificial intelligence.

We seek a versatile Senior Software Engineer who is passionate about performance optimization and generative AI. Our team builds software solutions that enable efficient inference on the latest and greatest generative AI models. We tackle problems on all levels of the stack—from server-level request batching to GPU kernel fusion—and collaborate with teams across diverse disciplines to push Nvidia's hardware to its full potential.

What You’ll Be Doing

• Cooperate with research teams to onboard new LLMs and VLMs into Nvidia's opensource AI runtimes
• Optimize inference workloads using sophisticated profiling and simulation tools
• Build SOLID, extendable inference software systems, and refine robust APIs
• Implement and debug low-level GPU code to harness the latest HW features
• Own end-to-end inference acceleration features and work with teams around the world to deliver production-grade products
  
  

• B.Sc., M.Sc. or equivalent experience in Computer Science or Computer Engineering
• 5+ years of relevant hands-on software engineering experience
• Profound knowledge of software design principles
• Strong proficiency in at least one system and one scripting language
• Strong grasp of machine learning concepts
• People person with excellent communication skills that enjoys collaboration and teamwork.
  
  

• Familiarity with Nvidia's DL software stack, e.g. Triton Inference Server, TensorRT-LLM, and Model Optimizer
• Proven track record of performance modeling, profiling, debugging, and development in a performance-critical setting with Nvidia's accelerators.
• Familiarity with LLM quantization, fine-tunning, and caching algorithms
• Proficiency in GPU kernel programming (CUDA or OpenCL)
• Prior experience working on a large software project with 50+ contributors