AI + Environment Summit 2025: Shaping the future of sustainable AI from Greater Zurich

AI for Biodiversity: Prof. Sara Beery (MIT) on species modeling

Prof. Dr. Sara Beery, Assistant Professor at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), develops computer vision methods for large-scale biodiversity and environmental monitoring. Her research focuses on making AI models resilient to domain shifts, imperfect data, and long-tailed species distributions – bridging ecological science and modern machine learning to support real-world conservation.

In her keynote, Beery introduced her team’s work on deep multimodal species occupancy modeling – a framework that merges camera-trap imagery, satellite data, and environmental covariates to estimate species presence more accurately. By accounting for real-world data complexity and detection uncertainty, the model moves beyond idealized lab settings to deliver insights that inform conservation policy, wildlife management, and habitat restoration.

💡Key takeaway:

Bringing AI into conservation means building models that can cope with nature’s messiness—integrating field data, uncertainty, and scale to deliver predictions robust enough for real ecological decisions.

Also Watch: Prof. Devis Tuia (EPFL / UZH) on Machine Learning for Earth

For a Swiss perspective, explore Prof. Dr. Devis Tuia’s keynote on “Machine Learning for Earth: Monitoring the Pulse of Our Planet with Sensor Data, from Your Phone All the Way to Space”. He shows how remote sensing data from satellites, drones, and ground sensors fuels machine learning (ML) models that transform raw signals into actionable insights for climate and biodiversity monitoring.
🎥Watch Prof. Tuia's keynote here. 

AI4EO for social good: Prof. Xiaoxiang Zhu (TUM) on remote sensing with ML

Prof. Dr. Xiaoxiang Zhu, Chair Professor for Data Science in Earth Observation at the Technical University of Munich (TUM), directs ML4Earth, the AI4EO Lab, and the Munich Data Science Institute. She develops remote sensing and machine learning (ML) methods that turn global Earth-observation data into actionable insights for urbanization, climate, and sustainable development.

In her keynote, Zhu shows how AI for Earth Observation (AI4EO) can serve society – from creating the Global Building Atlas, a worldwide 3D dataset of building footprints and heights, to tracking glacier retreat across decades of satellite imagery. She called for Earth-observation foundation models that are physics-informed, uncertainty-aware, and multi-sensor by design.

💡Key takeaway:

To deliver real social impact, AI for Earth Observation must go beyond accuracy – models need to be physically consistent, transparent, and scalable to turn remote-sensing data into trustworthy climate and urban insights.

AI to Impact – How environmental AI earns trust from lab to policy and markets

This closing panel discussion explored how AI can move beyond prediction to deliver measurable, equitable impact for climate, biodiversity, and society.

Panelists:

• Prof. Dr. Millie Chapman (ETH Zurich) 
  Assistant Professor of Environmental Policy at ETH Zurich. Her group works at the intersection of decision science, ecology, and data justice, critically exploring how computational tools can help meet biodiversity and climate targets. Formerly at Google Research, the National Center for Ecological Analysis and Synthesis (NCEAS), and Climate Change AI (CCAI), Chapman bridges AI, environmental policy, and ethics.
  🎥Watch her keynote: How AI can help inform just Environmental Futures

• Dr. Michael Anthony (Vertify.Earth)
  Co-founder and Executive Director of Vertify.Earth, developing AI tools for nature-positive interventions in fragile and climate-sensitive geographies. He advocates for transparency, local context, and trust between data producers and policy users.
  🎥 Watch his keynote: Context and Caution in Environmental AI

• Prof. Dr. S. Karthik Mukkavilli (Mercuria / KAIST)
  Group Head of AI at Mercuria Energy Group and Adjunct Professor at Korea Advanced Institute of Science and Technology (KAIST), focusing on foundation models, multimodal data, and reinforcement learning for Earth and energy systems. His work connects scientific modeling with real-world market deployment.
  🎥 Watch his keynote: AI for Complex Systems

The discussion revealed that building better models isn’t enough – impactful AI must integrate social, economic, and ethical realities. Chapman emphasized participatory design and fairness; Anthony stressed trust and data provenance amid political and regulatory tensions; Mukkavilli highlighted responsible deployment in high-stakes domains like energy trading and climate forecasting. 

💡 Key takeaway:

Environmental AI creates impact when it’s transparent, auditable, and co-designed with the people who use it – bridging lab insights with policy, markets, and conservation.