Poster at the AGU26 Annual Meeting. Beyond Scarce Labels: Foundation Models for Mapping Oil, Gas, and Power Infrastructure, with Ziheng Zhu and Da Zhang.
I build AI systems that turn climate evidence into policy intelligence.
My research asks when a satellite observation becomes evidence, and when evidence is strong enough to support a public decision. I combine remote sensing, machine learning and policy analysis to make methane emissions attributable and verifiable, and to assess climate scenarios transparently.

Poster at the AGU26 Annual Meeting. Beyond Scarce Labels: Foundation Models for Mapping Oil, Gas, and Power Infrastructure, with Ziheng Zhu and Da Zhang.
Paper accepted at IEEE BigData 2026. From Plume Events to Government Cases, Industry and Government Program.
Invited speaker, IIASA. Towards AI-Assisted Climate Scenario Assessment, AI for Climate Science Seminar Series.
I am a postdoctoral researcher at the Institute of Energy, Environment and Economy, within Tsinghua University’s Institute of Nuclear and New Energy Technology, as a Shuimu Tsinghua Scholar. My research concerns decision support for methane mitigation in the fossil fuel industry, connecting satellite observations, mathematical modelling and policy analysis.
I serve as a Chapter Scientist for the Working Group I contribution to the Seventh Assessment Report (AR7) of the Intergovernmental Panel on Climate Change (IPCC), in Chapter 5, Scenarios and projected future global temperatures.
From 2020 to 2024, I worked as a consultant with the United Nations Environment Programme, first on SDG-Meter, an AI tool that relates policy texts to the Sustainable Development Goals, and then with its Climate and Energy Branch on methane and climate intelligence.
I earned a PhD in Mathematics and Computer Science from Sorbonne University in 2024, with research conducted at the Centre for Applied Mathematics of École Polytechnique and at the Institut Supérieur d’Électronique de Paris (ISEP), in collaboration with UNEP. My doctoral research developed deep-learning methods that attribute satellite detections of methane plumes to oil and gas infrastructure.
I am also the founder and research lead of the Global Earth Intelligence Observatory (GEIO), an independent research initiative on how Earth observation evidence can responsibly inform public decisions.
I develop analytical systems for settings where scientific complexity, institutional accountability and real-world action meet, designed around the evidence standards, transparency and review processes that public institutions need.
How satellite detections of methane plumes become attributable, verifiable and administratively usable evidence: linking plumes to facilities, operators and public registers, following up alerts from UNEP’s Methane Alert and Response System, and clarifying what atmospheric evidence can establish under the EU Methane Regulation.
AI as a transparent analytical layer for exploring underexamined pathways, vetting scenario plausibility and coherence, and synthesising large ensembles for climate assessment processes such as those of the Intergovernmental Panel on Climate Change (IPCC).
Computer vision and vision-language models that map energy infrastructure from satellite imagery under scientific constraints: robustness, rigorous evaluation, and recognising when an image is insufficient and a model should abstain.
Postdoctoral Researcher, Institute of Energy, Environment and Economy
Institute of Nuclear and New Energy Technology
Chapter Scientist, Seventh Assessment Report (AR7), Working Group I: The Physical Science Basis
Chapter 5: Scenarios and projected future global temperatures
Founder & Research Lead, independent research initiative
globalearthintelligence.org
Consultant, Climate & Energy Branch (2022–2024)
Consultant in Innovative Technology (2020–2021)
PhD in Mathematics and Computer Science, in collaboration with UNEP
Research laboratories: Centre for Applied Mathematics, École Polytechnique & Institut Supérieur d’Électronique de Paris (ISEP)
Lecturer, engineering programme
Research Engineer, data science, signal and image processing (2019–2020)
Core Team Member (volunteer)
Summer school organisation, workshop and grant programme reviewing
Selected for Tsinghua University’s competitive postdoctoral fellowship programme supporting outstanding early-career researchers.
SDG-Meter, developed with the United Nations Environment Programme, ranked 9th in the Excellent Projects category of the Global Top 100 of AI solutions advancing the United Nations Sustainable Development Goals. International Research Centre on Artificial Intelligence under the auspices of UNESCO.