IBM Research is at the forefront of creating intelligent systems that transform how humans interact with computers. We are currently developing the next generation of configurable
Generalist AI Agents: CUGA [ https://cuga.dev https://github.com/cuga-project/cuga- agent]. This advanced technology is designed to independently navigate and execute tasks
across complex APIs MCP tools web environments and business workflows. CUGA currently ranks among the leaders in global benchmarks such as AppWorld and WebArena leveraging advancements in Agentic AI Large Language Models (LLMs) and Reinforcement Learning. It combines capabilities in natural language processing model fine-tuning and interactive decision-making. By integrating a semantic
understanding of APIs MCP tools and web elements with advanced action planning CUGA aims to handle tasks that typically require significant human intervention.
We are seeking a highly motivated intern to join our research team and help advance the
field of Generalist AI agents.
As a summer intern you will:
• Innovate & Develop: Design develop and refine cutting-edge AI agents and
Agentic frameworks incorporating advanced technologies such as LLMs
multimodal models and reinforcement learning systems.
• Experiment & Benchmark: Conduct rigorous experiments on leading
benchmarks—including WebArena AppWorld and Tue-bench—to assess and
enhance the capabilities of AI agents.
• Data & Prototyping: Curate datasets fine-tune models and demonstrate
performance capabilities through early-stage prototypes for web-based automation
solutions.
• Research Collaboration: Collaborate with cross-functional teams of researchers
and engineers to publish findings in top-tier venues and drive innovation within
IBM’s groundbreaking solutions.
Locations:
• IBM Research Lab Israel (Haifa University Campus)
• IBM Site Hashahar Tower Givataim (near Tel Aviv Arlozorov train station)
- Currently enrolled M.Sc. or Ph.D. student in Computer Science Information Systems or a related field with a focus on NLP Agentic systems and LLMs.
- Proven expertise in key domains such as LLMs/VLLMs Transformers Natural Language Processing (NLP) and Reinforcement Learning.
- Strong proficiency in Python development.
- Hands-on experience in implementing training and fine-tuning Transformers and LLMs.
Please include your grade sheet with your application.
- Solid understanding of AI agents and Agentic frameworks.
- Experience working with standard AI benchmarks.
- A publication track record in peer-reviewed conferences or journals is a significant- advantage.