PhD Research Intern, Efficient Deep Learning

What you'll do

• Research, design, and implement novel methods for efficient diffusion language models and multimodal generative models.
• Develop sampling efficiency, adaptive unmasking, self-speculation, and parallel decoding techniques.
• Build training and distillation pipelines for multimodal generation.
• Design hybrid inference orchestration for agentic AI across cloud and edge environments.
• Develop routing and scheduling policies for on-device and cloud expert delegation.
• Create resource-aware agent loops and adaptive routing systems.
• Publish original research in top-tier machine learning and computer vision venues.
• Transfer developed technologies to product groups for real-world application.

What we're looking for

• Currently pursuing a Ph.D. in Computer Science, Engineering, Electrical Engineering, or a related field.
• Excellent knowledge of the theory and practice of machine learning and deep learning.
• Experience with large language models, diffusion language models, multimodal/vision-language models, or agentic systems.
• Hands-on experience with large-scale model training, including data preparation and model parallelization (tensor and pipeline).
• Outstanding research track record with at least one top-tier conference publication (e.g., ICML, ICLR, NeurIPS, CVPR, ICCV).
• Excellent communication skills.
• Parallel programming experience, such as CUDA (preferred).
• Experience in hybrid cloud–edge inference, orchestration, adaptive routing, pruning, quantization, NAS, or efficient backbones (preferred).