Google DeepMind is inviting expressions of interest from qualified PhD researchers for a 2027 Student Researcher position focused on emergent capabilities of video models in Zurich, Switzerland. The research opportunity will explore how advanced video-generation and multimodal models develop and demonstrate capabilities, including visual intelligence, reasoning and other behaviours that may emerge as models become more sophisticated. The six-month, full-time position offers researchers an opportunity to contribute to an emerging area of artificial intelligence research while working with Google DeepMind researchers Robert Geirhos and Thaddäus Wiedemer. The role is paid and based on-site at the Google DeepMind office in Zurich. Interested candidates should note that the expression-of-interest form is an informal channel rather than an official Google DeepMind recruitment process. Applicants should also apply through the relevant official careers listing. Opportunity Overview Position: Student Researcher – Emergent Capabilities of Video Models Organisation: Google DeepMind Location: Zurich, Switzerland Work arrangement: Full-time, on-site Duration: Six months Start date: Between January and April 2027 Funding: Paid research position; exact compensation is not specified in the supplied announcement Research supervisors: Robert Geirhos and Thaddäus Wiedemer Expression-of-interest deadline: 20 October 2026 Official recruitment route: Google DeepMind Careers About the Research Project The project will investigate emergent capabilities in state-of-the-art video models, including models such as Gemini Omni. The central objective is to discover, evaluate and analyse capabilities that may not be explicitly programmed into a model but become apparent through experimentation and systematic testing. The research offers substantial scope for creative experimentation, coding and developing new approaches to understanding advanced AI systems. The precise research direction will be selected collaboratively, taking into account the interests and expertise of the researcher and supervisors. Previous research associated with Robert Geirhos includes the Physics-IQ benchmark, which has been used by organisations such as Google, NVIDIA, OpenAI, Meta and Black Forest Labs. Another example is Video Models Are Zero-Shot Learners and Reasoners, which attracted substantial public and media attention. These examples illustrate the potential for student research to produce influential benchmarks, new insights into model capabilities and research outputs relevant to the wider AI community. Required Qualifications and Experience Applicants must meet the core academic and technical requirements. Academic qualifications Currently enrolled in a PhD programme in machine learning, computer vision or a closely related field. A demonstrated record of first-author or joint-first-author publications at leading conferences or journals, such as NeurIPS, ICLR, ICML, CVPR or ICCV. An established research publication profile. The announcement identifies 100 or more Google Scholar citations as a useful benchmark, although this is not a strict requirement. Technical and research skills Practical experience using Python and deep learning frameworks. Knowledge of, or research experience with, generative models, large language models and video models. Ability to use agentic coding tools effectively while retaining strong manual coding skills. Creativity, research independence and the ability to develop promising research questions. Workshop papers do not count toward the specified first-author publication requirement at leading venues. Additional Skills That Could Strengthen an Application The following experience is desirable but not mandatory: Familiarity with diffusion models, in-context learning or psychophysics. Experience designing benchmarks and creative evaluations to investigate AI model capabilities. Experience designing and conducting human-subject experiments, either in laboratory settings or through crowdsourcing. Experience using artificial intelligence as a teaching tool. Candidates should present their research contributions clearly and provide evidence of their individual role in publications, experiments, software development and benchmark creation. Position Benefits and Research Experience The position offers six months of paid, full-time research experience in Zurich, with on-site collaboration at Google DeepMind. The researcher will work with experienced supervisors and contribute to an emerging field with opportunities for creative project development. The announcement also mentions free food at the office. The opportunity may be particularly valuable for PhD researchers interested in computer vision, multimodal learning, generative AI, video understanding, model evaluation and the scientific investigation of advanced AI capabilities. The announcement does not provide a specific salary figure, accommodation allowance or relocation package. Candidates should confirm compensation and any relocation or immigration support through the official recruitment process. How to Express Interest Candidates are asked to complete an online expression-of-interest form by 20 October 2026. The form requests the following information: Name and email address. Google Scholar profile and citation count. Number of peer-reviewed first-author or joint-first-author publications at leading venues. Personal website and GitHub or other code repository. Preferred start date. PhD start date and anticipated graduation date. University or research institution affiliation. Swiss or European Economic Area work authorisation status. A link to the applicant’s CV. A hypothetical research project title and abstract describing an idea they would like to explore. The form also allows applicants to provide optional details about their work permit status and any other relevant information. Important: The form explicitly warns applicants not to copy and paste an AI-generated response to the research-project question. Candidates should develop their own original idea and describe it in their own words. The purpose is to understand their research interests, creativity and approach to scientific problems. Work Authorisation Applicants are asked to indicate whether they are already authorised to work in Switzerland, authorised to work in at least one EEA country, or neither. Existing Swiss work authorisation is not stated as a universal prerequisite in the expression-of-interest form. The form indicates that Google typically makes efforts to support work-permit applications, and applicants are not required to already possess a Swiss work permit to express interest. However, work authorisation and immigration requirements should be confirmed through the official recruitment process. Official Application Links VISIT OFFICIAL WEBSITE TO APPLY Expression-of-interest form: Use the Google Form supplied in the original announcement. The expression-of-interest form is optional and does not replace the official recruitment process. Eligible researchers should review the official job listing, verify the application requirements and submit their application through the appropriate Google DeepMind careers channel. Final Note This opportunity is best suited to PhD researchers with a strong publication record, practical deep learning skills and a genuine interest in understanding the capabilities and limitations of advanced video models. Candidates who can demonstrate original research thinking, rigorous experimentation and strong technical implementation skills may be particularly well positioned to compete. For more opportunities such as these please follow us on Facebook, Instagram, Twitter, LinkedIn and WPChannel Disclaimer: Global South Opportunities (GSO) does not take ownership or credit for the opportunities shared on this platform. All opportunities are sourced from official announcements and external sources for informational purposes only. Applicants are encouraged to verify all details, deadlines, and eligibility requirements from the official source before applying. Opportunities may be updated or changed by the providing organisations.