Nature conservation faces a growing paradox: while unprecedented volumes of biodiversity data are being collected, effective conservation action is increasingly constrained by the lack of detailed, large-scale monitoring. In this talk I discuss how machine learning (ML) methods can help bridge this gap, by advancing the scale and resolution of biodiversity monitoring using multimodal data, collected by satellites, remote sensors, citizen-science and long-term monitoring studies. To demonstrate this, I will present two case studies from the United Kingdom: mapping the land cover of a National Park and predicting butterfly species occurrence. I will then discuss new challenges that arise when applying ML in ecology, including data requirements and interpretability, and outline how insights from other scientific disciplines can help address these challenges.