Senior Deep Learning Frameworks CUDA Software Engineer at NVIDIA

Department: Engineer, Sys SW

Compensation

$184,000 – $356,500/yr

Setup: Remote
Location: California · Austin, Texas · +1
Type: Full-time
Level: senior
Posted: Aug 31, 2026

Description

NVIDIA is hiring a Deep Learning Engineer to integrate CUDA features and runtime abstractions into AI frameworks, analyze and optimize AI workloads, improve the AI compiler-runtime interface, and design fault-tolerant and elastic large-scale systems. The role requires a computer science or related degree, at least 8 years of relevant experience, and expertise in deep learning frameworks, Python, C++, CUDA, performance profiling, and distributed machine learning. The base salary is 184,000 USD to 287,500 USD for Level 4 and 224,000 USD to 356,500 USD for Level 5, with equity and benefits; applications were accepted through September 4, 2026.

Required Skills

CUDA
PyTorch
JAX
TRT-LLM
vLLM
SGLang
Python
C++
Performance benchmarking
Distributed machine learning
AI compiler-runtime

Additional Information

Minimum Education: Bachelor's, master's, or PhD degree in Computer Science, Computer Engineering, Electrical Engineering, or a related field, or equivalent experience
Required Previous Experiences: At least 8 years of relevant industry experience or equivalent academic experience after completing the degree
Development experience with deep learning frameworks and inference engines
Rapid prototyping and development with Python, C++, CUDA, or related DSLs
Experience conducting performance benchmarking on AI clusters and familiarity with performance profiler toolchains
Understanding of HPC and AI communication concepts
Understanding of computer system architecture, hardware-software interactions, and operating systems principles
Preferred Qualifications: Deep expertise in the performance internals and execution graphs of major deep learning frameworks
Hands-on experience with CUDA, communication libraries, and distributed machine learning techniques
Expertise in training, distributed inference, MoE, reinforcement learning, or kernel authoring
Background in deep learning compilers, including graph-level and codegen approaches
Experience programming for compute and communication overlap in distributed runtime