Research Intern (Summer 2026)
Hybrid (San Francisco)$8k/mo - $9.3k/moInternPosted 11mo ago
About the role
Cua is building infrastructure that enables general-purpose AI agents to safely and scalably use real computers and applications. As a Research Intern, you’ll prototype, test, and benchmark multi-modal LLM-based agents, collaborate with engineers and researchers, and contribute to open-source research, benchmarks, and datasets.
Responsibilities
- Generate and curate large-scale, high-quality multi-modal data (GUIs, browsers, system UIs)
- Design and test single- and multi-agent systems for data and computer use
- Automate benchmarking of agent orchestration (with or without human-in-the-loop)
- Explore new training and inference techniques to boost reasoning and action-taking (e.g., RL-based agents)
- Develop benchmarks, tools, and datasets to evaluate agentic capabilities on Cua
- Collaborate with the founding team and contribute to research publications, open-source tools, and the broader community
Qualifications
- Currently a PhD student in Computer Science or related field (strong Master’s considered)
- Experience in applied research with a solid publication record
- Familiarity with modern multi-modal or reasoning agents (e.g., OS-Atlas, Qwen, GUI-R1)
- Hands-on experience with PyTorch, Python, and cloud compute (AWS, GCP, etc.)
- Comfortable designing experiments, evaluating models, and working with multi-modal data
- Excited by generative AI, agent systems, and pushing the boundaries of what’s possible
Skills
Required:PhD student in Computer Science or related fieldApplied research with publicationsMulti-modal or reasoning agents familiarityPyTorchPythonCloud compute (AWS, GCP, etc.)
Benefits
- Paid internship (3-month full-time preferred; part-time considered). Compensation will depend on location and experience
- Remote-friendly; SF-based team
- Learning environment – collaboration at the intersection of infrastructure and AI research
- Open-source visibility – contribute benchmarks and datasets used by the community
- Research impact – opportunity to publish, open-source, and influence open agent research
Last updated Sep 16, 2026. Interested? Apply on the company site.
About Cua
Docker for Computer-use Agents
Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).
open-source
More open roles at Cua
- Founding Technical GTM Lead1mo agoHybrid (San Francisco, CA, US) · GTM / Sales
- Founding Engineer, Infra & Agent Systems11mo agoHybrid (San Francisco, CA)