Agentic AI and Retrieval

The Agentic AI and Retrieval group studies how intelligent agents can search, reason with, and act on information in complex environments. We combine information retrieval, applied machine learning, and foundation models to build systems that perform multi-step web search, integrate evidence from heterogeneous sources, and coordinate with humans or other agents. A particular focus is agentic information retrieval: understanding how AI agents search differently from people and how retrieval systems should adapt to these new users. The group develops open, reproducible methods and infrastructure for reliable, efficient, transparent, and controllable AI, with web search serving as both a core research problem and a real-world testbed.

The Agentic AI and Retrieval group investigates how artificial intelligence systems find, evaluate, combine, and use information. Our work brings together three closely connected areas: agentic AI, applied machine learning, and web search. We develop AI agents for complex information-seeking and problem-solving tasks in which a system must plan several steps, formulate and reformulate queries, select tools, assess evidence, and decide when sufficient information has been gathered.

A central research question is how information retrieval changes when the user is an AI agent rather than a person. Agents can issue many queries, follow multiple reasoning paths, call specialised search services, and integrate evidence across documents and modalities. We study retrieval architectures for these settings and evaluate the complete search trajectory – including effectiveness, cost, latency, provenance, and failure modes – rather than only a single ranked result list.

Agentic Information Retrieval and Web Search: Developing retrieval systems for agents that autonomously search, reformulate queries, compare sources, and use information across multiple steps. Topics include multi-hop retrieval, retrieval-augmented reasoning, ranking and re-ranking, the role of embeddings, and evidence-grounded search.

Applied Machine Learning: Using statistical learning, neural models, foundation models, and language models to analyse heterogeneous real-world data and build systems for search, classification, prediction, and decision support. Particular interests include robust learning with noisy or incomplete information, multimodal integration, and practical evaluation.

Multi-Agent and Human-Agent Coordination: Studying how specialised agents, foundation models, human experts, and existing information systems collaborate. We ask when coordination improves problem solving, where collective systems fail, and how responsibilities can be distributed between agents and people.

Open Search Infrastructure: Building and using open, inspectable web-search infrastructure as a research platform. Work around OpenWebSearch.eu, the Open Web Index, and agent-facing services such as OURRS enables reproducible experiments at realistic scale and supports new search and AI applications.

Reliable and Efficient Agents: Improving robustness, transparency, controllability, scalability, and resource efficiency. We investigate how retrieval quality, uncertainty, cost, and latency accumulate across multi-step agentic workflows and how users can inspect and control those processes.

Potential Projects

Potential research, student, and infrastructure projects connected to the Agentic AI and Retrieval group: