Job description

Are you passionate about pushing the limits of real-time large language model inference? Join NVIDIA’s TensorRT Edge-LLM team and help shape the next generation of edge AI for automotive and robotics. We build the software stack that enables Large Language, Vision-Language, and Multimodal (LLM/VLM/VLA) models to run efficiently on embedded and edge platforms — delivering cutting-edge generative AI experiences directly on-device. What you’ll be doing: • Develop and evolve a state-of-the-art inference framework in modern C++ that extends TensorRT with autoregressive model serving capabilities, including speculative decoding, LoRA, MoE, and KV cache management. • Design and implement compiler and runtime optimizations tailored for transformer-based models running on constrained, real-time platforms. • Collaborate with teams across CUDA, kernel libraries, compilers, and robotics to deliver high-performance, production-ready solutions. • Contribute to CUDA kernel and operator development for critical transformer components such as attention, GEMM, and MoE. • Benchmark, profile, and optimize inference performance across diverse embedded and automotive environments. • Stay ahead of the rapidly evolving LLM/VLM ecosystem and bring emerging techniques into product-grade software. What we need to see: • BS, MS, PhD, or equivalent experience in Computer Science, Electrical/Computer Engineering, or a closely related field. • 4+ years of relevant software development experience. • Deep understanding of transformer models and inference optimization techniques (e.g., quantization, tensor parallelism, or memory-efficient scheduling). • Proficient programming ability with modern C++ (C++11/14/17 and beyond). • Familiarity with popular LLM frameworks and libraries such as TensorRT, TensorRT-LLM, vLLM, SGLang, MLC-LLM, or FlashInfer. • A track record of strong software design, execution, and collaboration across fields. Ways to stand out from the crowd: • Demonstrated development experience or open-source contributions to LLM inference frameworks and libraries, such as SGLang, vLLM, or FlashInfer. • Proficiency with CUDA, including efficient kernel development, performance profiling, and GPU architecture fundamentals. • Prior work on autoregressive LLM serving systems, including speculative decoding or KV cache management. • Familiarity with compiler infrastructure for large language model inference. • Exposure to robotics or embedded AI pipelines, including optimizing for low-latency, resource-constrained systems. NVIDIA is widely considered to be one of the technology world’s most desirable employers. We hire some of the most brilliant and forward-thinking people in the world. If you thrive on innovation, autonomy, and technical excellence, come join us to shape the future of edge AI. #LI-Hybrid Your base salary will be determined based on your location, experience, and the pay of employees in similar positions. The base salary range is 152,000 USD - 241,500 USD for Level 3, and 184,000 USD - 287,500 USD for Level 4. You will also be eligible for equity and benefits . Applications for this job will be accepted at least until August 9, 2026. This posting is for an existing vacancy. NVIDIA uses AI tools in its recruiting processes. NVIDIA is committed to fostering an inclusive work environment and proud to be an equal opportunity employer. As we highly value diversity in our current and future employees, we do not discriminate (including in our hiring and promotion practices) on the basis of race, religion, color, national origin, gender, gender expression, sexual orientation, age, marital status, veteran status, disability status or any other characteristic protected by law.

Requirements

Must-haves

Develop and evolve a state-of-the-art inference framework in modern C++ that extends TensorRT with autoregressive model serving capabilities, including speculative decoding, LoRA, MoE, and KV cache management.
Design and implement compiler and runtime optimizations tailored for transformer-based models running on constrained, real-time platforms.
Collaborate with teams across CUDA, kernel libraries, compilers, and robotics to deliver high-performance, production-ready solutions.
Contribute to CUDA kernel and operator development for critical transformer components such as attention, GEMM, and MoE.
Benchmark, profile, and optimize inference performance across diverse embedded and automotive environments.
Stay ahead of the rapidly evolving LLM/VLM ecosystem and bring emerging techniques into product-grade software.
BS, MS, PhD, or equivalent experience in Computer Science, Electrical/Computer Engineering, or a closely related field.
4+ years of relevant software development experience.
Deep understanding of transformer models and inference optimization techniques (e.g., quantization, tensor parallelism, or memory-efficient scheduling).
Proficient programming ability with modern C++ (C++11/14/17 and beyond).
Familiarity with popular LLM frameworks and libraries such as TensorRT, TensorRT-LLM, vLLM, SGLang, MLC-LLM, or FlashInfer.
A track record of strong software design, execution, and collaboration across fields.
Demonstrated development experience or open-source contributions to LLM inference frameworks and libraries, such as SGLang, vLLM, or FlashInfer.
Proficiency with CUDA, including efficient kernel development, performance profiling, and GPU architecture fundamentals.
Prior work on autoregressive LLM serving systems, including speculative decoding or KV cache management.
Familiarity with compiler infrastructure for large language model inference.
Exposure to robotics or embedded AI pipelines, including optimizing for low-latency, resource-constrained systems.
You will also be eligible for equity and benefits .
Applications for this job will be accepted at least until August 9, 2026.
This posting is for an existing vacancy.
NVIDIA uses AI tools in its recruiting processes.

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

llmai

Platform: eightfold
Effort: medium

Benefits

Benefits detected in the listing.

Vision