Agentic AI for EO

As AI systems increasingly move from passive tools to autonomous agents capable of taking action in the world, a new paradigm is emerging that transforms both software development and scientific inquiry. The exponential growth of data and computational power, combined with the rise of LLMs as sophisticated reasoning engines, has initiated a shift toward agentic workflows, where LLMs and other computational tools are orchestrated as autonomous or semi-autonomous agents to tackle complex, multi-step problems.

This session explores both dimensions of this paradigm shift. It begins with a practical exploration of what it means to build and operate software in the age of agentic computing, using the Blablador LLM inference service as a case study. Through real-world examples, e.g., mapping illegal landing strips in the Amazon, the session shows how agentic systems can change the mindset of the practitioner, and offers practical takeaways and lessons from the edge of what is currently possible.

The session then turns to the application of agentic workflows for scientific discovery. When applied to scientific inquiry, this approach facilitates automating hypothesis generation, data analysis, and insight extraction. However, unique challenges persist in applying these agentic systems to specialized domains like Earth Observation, particularly in grounding LLMs, reasoning with geospatial datasets, and ensuring the reliability of autonomous systems. Participants will be guided through the lifecycle of designing and implementing agentic workflows for scientific discovery using a new framework called Accelerated Knowledge Discovery.

Agenda

Block 1: Agentic computing and Blablador

• Building and operating software in the age of agentic AI
• Case study: vibe coding a downloader for mapping illegal landing strips in the Amazon

Block 2, 2 and 3 Agentic workflows for scientific discovery

• Hands-on: designing and implementing agentic workflows for EO using Accelerated Knowledge Discovery