Location: Mountain View, CA
Cap-Exempt H-1B Position — No lottery required
**Lab Summary:** Samsung AI Research Center (SAIC) located in Mountain View, California, is currently recruiting outstanding research interns for the Vision Intelligence lab. Our goal is to perform long-term research with potential for direct impact on future Samsung products reaching hundreds of millions of users worldwide. We are focused on pushing the state-of-the-art in multimodal understanding and generation. **Position Summary:** In this position, the student will collaborate with leading researchers on a project related to the lab's interests in multimodal (image, video, 3D, text) retrieval, question answering, reasoning, generation and editing. **Position Responsibilities:** - Conduct research on multi-modal models in one the above-mentioned area - Write maintainable code to implement research ideas - Work closely with the team to discuss findings and improve the method - Executing experiments, documenting their progress and presenting insights - Prepare findings for submission to a major conference or journal **Required Skills:** - Currently pursuing a MS or PhD degree - Experience with programming in Python, Pytorch, Transformers etc. - Previous publications in top-tier conference and journals in CV/ML/AI such as CVPR, ECCV, ICCV, ICML, NeurIPS, AAAI - Experience in one or more of the following areas: - Experience in VLMs including model architecture design, pretraining or post training (SFT, RLHF, GRPO) techniques - Experience in diffusion models and other generative AI approaches for image/video/3D domains - Experience in design and development of efficient multi-modal models using techniques like knowledge distillation, mixture of experts etc. - Experience in Vision Language Action Models (VLA) for Agentic systems in robotics, GUI based task automation etc. - Strong oral and written communication skill **Additional Information** Disclosure of Trade Secrets Samsung has a strict policy on trade secrets. In applying to Samsung and progressing through the r