Cmbagent Lab @ Cambridge

Boris Bolliet

Boris Bolliet Boris' research focuses mainly on the development of AI systems that can autonomously conduct scientific research. He is the creator of Cmbagent, the award-winning first multi-agent system applied to Cosmological data analysis. He is also one of the three creators of Denario, an end-to-end framework for the automated and modular production and execution of scientific research projects across multiple domains, from Physics to Neuroscience. Boris leads agentic AI research in the Infosys-Cambridge AI Centre.

Current Positions

Previous Positions

  • Assistant Teaching Professor in Data Intensive Science (Cambridge, 2024-2026)
  • Assistant Research Professor (DAMTP, Cambridge, 2022-2024)
  • Guest Researcher (Flatiron Institute, CCA, New York, 2021-2023)
  • Research Associate (Columbia University, PI: Colin Hill, 2020-2022)
  • Research Associate (JBCA, Manchester, PI: Jens Chluba, 2017-2020)
  • Fulbright Visiting Student Researcher (Louisiana State University, 2016)
  • PhD (Grenoble, France, PI: Aurélien Barrau, 2013-2017)

Current Group Members

Previous Group Members

  • Celia Lecat (MSc, 2025) → Master MVA, Paris
  • Jamie Martin (MSc, 2024-2025) → ML Engineer at Neutreeno
  • Kahaan Gandhi (BSc, 2025) → Caltech
  • Xueqing Xu (MPhil, 2025) → AI Engineer at Barclays
  • Adrian Dimitrov (Part III, 2024-2025) → Full Stack Developer at Jaid
  • Edwin Robinson (MSc, 2024-2025)

Research Focus

We develop AI systems that can autonomously conduct scientific research.

Our research is funded by:

Selected Publications and Preprints

  • T. Borrett, L. Xu, A. Nilipour, B. Bolliet et al., Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research, arXiv:2604.09621 - accepted, Communications AI and Computing (Nature Portfolio)
  • X. Xu, B. Bolliet, A. Dimitrov et al., Evaluating Retrieval-Augmented Generation Agents for Autonomous Scientific Discovery in Astrophysics, arXiv:2507.07155 - spotlight, ML4Astro at ICML 2025
  • F. Villaescusa-Navarro, B. Bolliet, P. Villanueva-Domingo et al., The Denario project: Deep knowledge AI agents for scientific discovery, arXiv:2510.26887 - accepted, PRX
  • L. Xu, M. Sarkar, A. I. Lonappan, Í. Zubeldia et al., Open Source Planning & Control System with Language Agents for Autonomous Scientific Discovery, arXiv:2507.07257 - the cmbagent system paper

Full list: INSPIRE-HEP and arXiv.

Research Software

  • DENARIO — Modular Automation of Scientific Research with Multi-Agent Systems
  • CMBAGENT — Autonomous Research Backend for AI Agents
  • SKEPTHICAL — The AI Reviewer
Our group is involved in the following courses:

  • A8 - Agentic AI and Reinforcement Learning - MPhil in Data Intensive Science.
  • C1 - Research Computing and Software Development - MPhil in Data Intensive Science, MPhil in Economics and Data Science, and the Centre for Doctoral Training in Data Intensive Science (CDT). Extending to Computational Biology (Maths) and Planetary Science and Life in the Universe (Institute of Astronomy) from Michaelmas 2026. Part of the course material is published openly at researchcomputing.readthedocs.io.

University-wide events

We welcome top students and postdocs interested in working with us to get in touch. Note that due to the high number of emails we receive (thank you so much!), we truly apologise for not being able to answer all of them.

E-mail: bb667@cam.ac.uk

Address: Cavendish Laboratory, JJ Thomson Avenue, Cambridge CB3 0HE, United Kingdom

Office: Room F26 in Battcock