Deep Learning Scientist Intern, LLM, Multimodal and Agentic AI - Summer 2027

NVIDIA•Santa Clara, CA

About The Position

As a Deep Learning Scientist Intern, you will research, design, and develop large language models, multimodal LLMs, and agentic AI systems. You will apply machine learning and deep learning techniques across language, vision, speech, audio, and video to develop and improve next-generation foundation models. The work includes developing, training, fine-tuning, and evaluating models; exploring model capabilities such as reasoning, multimodal understanding, planning, tool use, conversational interaction, and multi-step task completion; and building datasets and evaluation benchmarks. You will prototype new architectures, algorithms, training strategies, and inference techniques, and collaborate with researchers and engineers to scale successful ideas. This is a summer 2027 internship. The expected hourly rate is 20 USD–94 USD, and interns are eligible for Intern benefits.

Requirements

Pursuing an MS or PhD in Computer Science, Electrical Engineering, Artificial Intelligence, Applied Mathematics, or a related field.
Strong knowledge of machine learning and deep learning techniques, algorithms, and modern foundation model architectures, including Transformers.
Experience or exposure to large language models, vision-language models, or multimodal models, with an emphasis on reinforcement learning for agentic reasoning, planning, tool use, and coding.
Excellent Python programming skills.
Experience with deep learning frameworks such as PyTorch, NeMo, Megatron-LM, TensorFlow, or related frameworks.
Experience developing, training, fine-tuning, or evaluating deep learning models, including familiarity with data processing, dataset development, experiment tracking, and model evaluation workflows.
Strong problem-solving skills and demonstrated ability to rapidly prototype, experiment, and evaluate new ideas.

Nice To Haves

Experience with large language models, multimodal foundation models, vision-language models, or agentic AI research and development.
Background with post-training techniques such as instruction tuning, preference optimization, RLHF, RLVR, GRPO, or related reinforcement learning methods.
Experience with multimodal reasoning across text, image, video, speech, or audio, including conversational or audio-language models.
Experience with reasoning, planning, tool use, function calling, or multi-step agentic systems.
Background with large-scale model training, distributed frameworks such as NeMo or Megatron-LM, GPU computing, or inference technologies such as vLLM or TensorRT-LLM.

Responsibilities

Research and develop large language models, multimodal foundation models, and agentic AI systems across text, vision, speech, audio, and video.
Develop, train, fine-tune, and evaluate foundation models using supervised fine-tuning, instruction tuning, preference optimization, reinforcement learning, and related post-training techniques.
Develop and evaluate capabilities including reasoning, multimodal understanding, planning, tool use, conversational interaction, and multi-step task completion.
Build and curate datasets and evaluation benchmarks for model training, post-training, and performance analysis.
Prototype model architectures, algorithms, training strategies, and inference techniques.
Collaborate with researchers and engineers to scale successful ideas.