About the role
NVIDIA is searching for an outstanding PhD intern working on efficient deep learning to join the Deep Learning Efficiency Research (DLER) team. We are passionate about research that pushes boundaries but also has impact in the real world. The team has two core focuses: (1) efficient diffusion language models and multimodal generative models, and (2) efficient agentic AI with hybrid inference orchestration across cloud and edge. We are also excited about post-training model optimization (pruning, quantization, NAS), efficient architecture design, adaptive/dynamic inference, and resource-efficient training and finetuning.
You will work within an amazing and collaborative research team that consistently publishes at the top venues in computer vision and machine learning. Our existing expertise includes computer vision, deep learning, generative models, diffusion LLMs, multimodal models, and hybrid cloud–edge agentic systems. Your contributions have the chance to create real impact on our products.
What you'll be doing:
Research, design, and implement novel methods for efficient deep learning in one or both of the team’s focus areas:
Diffusion LLMs and multimodal models — sampling efficiency, adaptive unmasking, self-speculation / parallel decoding, training and distillation pipelines, and multimodal generation.
Efficient agentic AI — hybrid inference orchestration across cloud and edge, routing and scheduling policies, on-device vs. cloud expert delegation, and resource-aware agent loops.
Publish original research.
Collaborate with other team members and teams.
Work with product groups to transfer technology.
Collaborate with external researchers.
What we need to see:
Pursuing a Ph.D. in Computer Science/Engineering, Electrical Engineering, etc.
Excellent knowledge of theory and practice of machine learning and deep learning.
Experience with large language models, diffusion language models, multimodal / vision-language models, or agentic systems is required.
Hands-on experience with large-scale model training including data preparation and model parallelization (tensor and pipeline) is required.
Outstanding research track record with at least one top-tier conference (ICML, ICLR, NeurIPS, CVPR, ICCV, etc.).
Excellent communication skills.
Ways to stand out from the crowd:
Parallel programming (e.g., CUDA).
Interest or experience in hybrid cloud–edge inference, orchestration, or adaptive routing.
Background in pruning, quantization, NAS, or efficient backbones.
NVIDIA is widely considered to be one of the technology world’s most desirable employers with competitive salaries and a generous benefits package, we have some of the most forward-thinking and hardworking people in the world working for us. And, due to unprecedented growth, our best-in-class engineering teams are rapidly growing. If you're a creative and autonomous engineer with a real passion for computer architecture and technology, we want to hear from you!
Our internship hourly rates are a standard pay based on the position, your location, year in school, degree, and experience. The hourly rate for our interns is 38 USD - 94 USD.You will also be eligible for Internbenefits.
Applications for this job will be accepted at least until October 9, 2026.This posting is for an existing vacancy.
NVIDIA uses AI tools in its recruiting processes.
NVIDIA is committed to fostering an inclusive work environment and proud to be an equal opportunity employer. As we highly value diversity in our current and future employees, we do not discriminate (including in our hiring and promotion practices) on the basis of race, religion, color, national origin, gender, gender expression, sexual orientation, age, marital status, veteran status, disability status or any other characteristic protected by law.Originally posted on Himalayas