Build observability across agent runs: what the agent did, where it failed, and why.
Champion guidelines for secure and reliable agent workflows, including data handling, access control, and interaction boundaries.
Serve as a key technical resource for solving sophisticated integration issues between agents and internal tooling.
What we need to see
B.Sc. or above in Computer Science, Computer Engineering, or a related field, or equivalent experience.
5+ years of hands-on experience in software engineering, with demonstrated ownership of production systems from design through deployment.
Expert-level programming skills in C++, with strong Python skills alongside it.
Strong understanding of the full stack, including hardware: memory, I/O, networking, accelerators, and where real performance bottlenecks occur.
Current, practical knowledge of how to build systems around AI models: agent loops, tool interfaces, context retrieval and management, and common failure modes.
Understanding of inference serving, including request lifecycle, batching, caching, and the tradeoffs between throughput, latency, and cost.
Ways to stand out from the crowd
Experience writing hardware simulation software — network, system, or architectural simulators, models, or testbenches.
Networking experience — protocols, fabrics, switching, or RDMA — and experience working alongside silicon, systems, or architecture teams.
Hands-on experience with inference serving engines such as vLLM, TensorRT-LLM, or Triton Inference Server, including low-level internals such as KV cache, batching and scheduling, and quantization, and related performance work such as profiling and GPU programming.
Hands-on experience building or fine-tuning LLMs or other generative models.
Agent workflows, tooling, or context pipelines adopted by other engineering teams.
NVIDIA is home to some of the most innovative and dedicated professionals in the industry. We are committed to fostering a diverse work environment and are proud to be an equal-opportunity employer.