Research Scientist: Human-Centric Visual Intelligence

Job Number: P25 F12

Honda Research Institute USA (HRI-US) is seeking a Research Scientist to advance the state of the art in human-centric visual intelligence. This role focuses on developing AI systems that perceive, interpret, and reason about human activities, interactions, intentions, and experiences across vision, language, and action. The successful candidate will conduct fundamental research in multimodal learning and human understanding, developing novel algorithms and models that enable AI systems to better understand people and their behavior in complex real-world environments.

San Jose, CA

• Conduct original research, develop novel methods and advance state of the art to enable AI systems to perceive, understand, and reason about human activities, behaviors, interactions, and intentions.
• Investigate multimodal representation learning, alignment, temporal reasoning, and long-horizon understanding of human activities and experiences.
• Develop methods for modeling latent human states, such as intent, goals, beliefs, attention, and other unobservable factors that influence behavior.
• Design and execute rigorous experiments, benchmarking studies, and ablation analyses to validate research hypotheses.
• Contribute to design, collection and annotation of new impactful video datasets.
• Collaborate with interdisciplinary teams of scientists and engineers to translate research advances into impactful AI systems.
• Publish research findings at leading AI, machine learning, and computer vision conferences and journals.
• Contribute to research strategy and intellectual property development.

• Ph.D. in Computer Science, Electrical Engineering, Cognitive Science, Computational Neuroscience, or a related field.
• Strong publication record at leading AI, machine learning, or computer vision venues such as CVPR, ICCV, ECCV, NeurIPS, ICLR, AAAI, or equivalent.
• Expertise in generative models and latent-variable learning, including Variational Autoencoders (VAEs), Diffusion Models, Generative Adversarial Networks (GANs), or related approaches.
• Strong foundation in machine learning, deep learning, and modern AI methodologies.
• Research experience in video understanding, activity recognition, and temporal modeling.
• Expertise in multimodal representation learning and alignment, vision and text encoders, and large-scale video datasets.
• Excellent communication, presentation, and collaboration skills.
• 1 - 3 years of relevant work experience.

• Research expertise in long-range action understanding, including action segmentation, temporal alignment, action anticipation and other long-horizon activity modeling.
• Experience developing methods for inferring human activities, intentions, or future actions from incomplete, ambiguous, or partially observed data.
• Experience with human pose estimation, hand pose estimation, or their application to activity understanding.
• Research background of multimodal models, including Vision-Language Models (VLMs), Multimodal Large Language Models (MLLMs), and related architectures.
• Understanding of multimodal training methodologies, architectures, adapters, objectives, and data curation strategies.
• Experience in modeling latent human states, intentions, goals, beliefs, attention, or other unobservable factors that influence behavior.
• Demonstrated ability to initiate and lead impactful research projects.

Desired

Start Date:

7/6/2026

Position