Research Intern – Multimodal Foundation Model for… at SonyAI · Remote

Summary: Sony AI America is a research organization focused on leveraging AI to enhance human creativity. The research intern will develop methodologies and prototype solutions for next-generation foundation models for vision, collaborating with scientists and engineers to solve complex challenges in the field. Responsibilities: - Conduct fundamental and innovative development in low-cost yet powerful vision-language models (VLM), unified models, automatic model compression, optimization and deployement on cloud and edge - Design or implement state-of-the-art techs on model compression, inference speedup, deployement on harwares, tool automation - PoC for various vision+text, generation relevant tasks (VQA, captioning, understanding, etc) and hardwares - Contribute to library and tool development to support business; or Publish influential research in top-tier conferences and journals Required Qualifications: - Currently has, or is in the process of obtaining, a master/PhD degree in computer science or related field - Be very self-motivated and capable of proposing and implementing innovative ideas - Solid presentation and communication skills to internal and external audiences - Publications or expertise in compact foundation model development and deployment. Influential open-source projects or paper publication at top conferences, e.g., CVPR, ICCV, ECCV, NeurIPS, ICML, ACL, etc - Solid coding skills in Python, Pytorch, etc Preferred Qualifications: - Better to have front-end development experience Required Skills: Python, Pytorch, Foundation model development Important Skills: Front-end development Nice-to-Have Skills: Self-motivated, Presentation skills, Communication skills