Senior Software Engineer - VLM Microservices for Neural Reconstruction

NVIDIA • Redmond, WA

14h

About The Position

NVIDIA is a world-leader in Gaussian Splatting and Neural reconstruction. Our team builds the Omniverse NuRec SDK to enable robotic, healthcare, and AV developers to build better models faster with closed-loop validation and closed-loop training grounded in real-world scenarios. Are you fascinated by the latest developments in Vision Language Models and 3D reconstruction? We are looking for a driven Software Engineer to bring ground breaking models into NVIDIA's software ecosystem for neural graphics to improve its robustness and accuracy and enrich its capabilities. Your software will help bridge the gap between the real world and simulations. Join our amazing team and help us build the future of Physical AI!

Requirements

• Master's of Science in Computer Science + 3 years, or Electrical Engineering, Bachelor of Science (or equivalent experience) + 5 years of experience.
• History of building, validating and releasing production-grade AI distributed systems, backend services, microservices, and cloud technologies.
• Deep technical expertise in distributed applications using Docker, Kubernetes, endpoints and their APIs (REST, gRPC), Helm.
• Hands-on experience with modern inference platforms (vLLM, SGLang, Torch, TRT, TRT-LLM).
• Proficiency with Python and C++.
• Excellent software engineering fundamentals (source control, CI/CD, testing/validation, packaging, containerization).
• Excellent written, visual, and verbal communication.
• Curiosity and drive to learn new technologies and partner across teams and functions.

Nice To Haves

• Track record contributing to open-source or production-grade software. Are you a vLLM contributor already?
• Experience with ML model engineering: training, fine-tuning, distillation, quantization.
• Experience with low-level optimization of ML models (CUDA kernels)
• Strong fundamentals in 3D graphics, 3D computer vision or neural reconstruction (NERFs, Gaussian Splats).
• History of multidisciplinary creativity and innovation around software engineering in multiple problem domains.

Responsibilities

• Design, build, and optimize containerized inference execution for the latest 3D VLMs from NVIDIA, turning research work into production-grade, highly optimized software (NIMs, NVIDIA Inference Microservices)
• Develop benchmarks to validate the models accuracy and performance (latency, throughput, scalability)
• Release and maintain the models and their pipelines throughout their lifecycle (bug fixes, security patches)
• Contribute VLM-related features to Open-Source projects like vLLM
• Collaborate closely with Research and Product teams and influence our common roadmaps

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level