Research and Engineering Intern, Agentic Systems Intelligence – Summer 2027

NVIDIA•Santa Clara, CA

About The Position

NVIDIA’s Agentic Systems Intelligence (ASI) team is seeking a Research and Engineering Intern to advance high-performance AI and build next-generation agentic AI systems for the CUDA ecosystem. The team works across the full stack, including model training and improvement, agent architectures and multi-agent systems, and the systems, runtimes, and evaluation frameworks needed to make agents effective at scale. The intern will investigate open-ended research questions and turn promising ideas into working systems that reason about, generate, optimize, and operate across NVIDIA’s accelerated computing stack. Work may focus on model and agent capabilities, scalable agent systems and runtimes, or high-fidelity training and evaluation environments. The intern will collaborate with NVIDIA software, model, and hardware teams, with opportunities to bring new capabilities into NVIDIA products.

Requirements

Currently pursuing a Master’s or PhD in Computer Science, Electrical Engineering, or a similar field.
Strong foundations in one or more of: AI systems, agentic AI, machine learning, reinforcement learning, high-performance computing, or computer architecture.
Hands-on research or development experience in agent systems, coding agents, model training, reinforcement learning, or modern AI inference systems.
Strong Python programming skills and proficiency in C/C++ or Rust, with solid software engineering fundamentals.
Experience with GPU computing and an interest in performance optimization and working across system boundaries.

Nice To Haves

PhD research in a relevant area, with demonstrated ability to formulate research questions, build working implementations, and evaluate them rigorously.
Experience building coding agents, autonomous software-engineering systems, or agent training and evaluation environments.
Hands-on experience with GPU programming and performance optimization, or systems such as TRT-LLM, SGLang, or vLLM.
Evidence of technical depth and ownership through impactful research prototypes, deployed systems, or substantial open-source contributions.
Relevant publications; strong engineering and measurable real-world impact are especially valued.

Responsibilities

Research, build, and optimize agentic AI systems for the CUDA ecosystem, including agent architectures, multi-agent workflows, tool use, memory, and orchestration.
Develop data pipelines, environments, verification tools, reward mechanisms, and evaluation frameworks to improve model and agent capabilities through rigorous experimentation.
Co-design and optimize agentic systems across NVIDIA’s software and hardware stack, connecting advances in models and agents with compilers, runtimes, libraries, kernels, and GPUs.
Demonstrate measurable improvements in correctness, performance, or developer productivity.