LLM Engineer, Agentic Researcher Platform

NVIDIA is seeking a Machine Learning Engineer to help build the core of an autonomous, agentic platform that optimizes machine-learning models end-to-end. The platform optimizes the model itself (architecture, hyperparameters) and its implementation (the CUDA/Triton code it compiles to) across domains.

Responsibilities:

• Develop and advance a self-governing, agentic platform that optimizes AI models end-to-end , architecture, hyperparameters, and the GPU code they compile to.
• Leverage AI-native and agentic workflows to accelerate research, experimentation, evaluation, and deployment of AI systems.
• Establish and drive benchmarking frameworks that measure accuracy, latency, memory footprint, throughput, and cost , including head-to-head comparisons that prove the agent beats existing automated search.
• Design and deploy with strong consideration for reproducibility, AI safety, sandboxing, and compute-cost governance.
• Lead technical initiatives, mentor engineers, and foster a One Team culture through close collaboration across research, engineering, and product teams.

Requirements:

• Master's degree in Computer Science, AI, Electrical Engineering, or equivalent experience.
• 3+ years of experience building and deploying ML, LLM, or model-optimization systems.
• Strong Python skills and hands-on experience with PyTorch (or TensorFlow).
• Hands-on experience with automated experimentation , hyperparameter optimization, AutoML, or NAS.
• Experience building LLM-agent systems (reasoning, tool use, multi-step orchestration) and/or production ML pipelines and MLOps infrastructure.
• Proven technical leadership and mentoring experience, and strong problem-solving, communication, and teamwork skills.

Preferred qualifications:

• Hands-on experience with NVIDIA AI technologies such as NeMo, TAO, Triton, CUDA, NIM, and Nemotron.
• Experience building agentic AI systems with reasoning, tool use, and code generation.
• Expertise in optimization: evolutionary and quality-diversity search (e.g. MAP-Elites), Bayesian optimization, and multi-fidelity methods (Hyperband/ASHA).
• GPU performance work , CUDA/Triton kernels, torch.compile, operator fusion, quantization , and interest in inference-efficiency domains such as AI-RAN.
• Experience benchmarking AI systems for accuracy, latency, memory, reliability, and cost. A research track record (publications or credible reproductions) in AutoML, NAS, LLM agents, or optimization.