Senior Systems Software Engineer, Machine Learning

NVIDIA•Durham, CA

About The Position

NVIDIA is transforming computer graphics, PC gaming, accelerated computing, and machine learning. We are tapping into the unlimited potential of AI to define the next era of computing. This role involves working on generative AI, LLMs/VLMs, computer vision, and agentic systems. The team is a mix of engineers and scientists who value rigor and timely delivery. The culture is collaborative, low-ego, and built around ownership, focusing on building systems that get used and making data-driven decisions.

Requirements

Masters degree, or preferably a PhD degree in Computer Science or a related field or equivalent experience.
5+ years of experience.
Solid mathematical and algorithmic foundation and proven expertise demonstrated through research publications, internships, or significant project experience.
Strong background in computer vision and deep learning.
Excellent programming skills in Python and C/C++.
Excellent software engineering fundamentals.
Ability to develop code in Unix/Linux environments.
Excellent written, visual, and verbal communication skills to present performance challenges, tradeoffs, and architectural alternatives.
Strong collaboration skills to partner with other teams.

Nice To Haves

Experience designing and operating multi-agent pipelines in production, including handling non-deterministic failures, retry logic, and tool-call error recovery.
Shipped a product feature backed by a VLM (e.g., image captioning, document understanding) — including handling inference latency, cost-per-call tradeoffs, and degraded-mode fallbacks.
Shipped AI-powered features to real users — not just prototyped with agent frameworks.

Responsibilities

Convert research into real products.
Help build workflows that diversify datasets and/or populate data.
Ship machine learning workflows/pipelines fast and iterate faster.
Leverage LLM/VLM and agents in the data generation pipeline.
Define evaluation criteria and run offline evaluations before any model or prompt change reaches production.