This year, the school is explicitly structured around a top-down and bottom-up integration of AI in microscopy:

  • Top-down: Agentic AI for instrumentation — how to design AI agents that interact with microscopes through APIs and orchestrators, enabling remote operation, cloudified workflows, and federated experimental systems.
  • Bottom-up: Reward-driven workflows — how to build decision-making frameworks grounded in physics and experimental feedback, enabling robust, adaptive, and autonomous exploration.

The school will cover core ML concepts, but the primary focus is on next-generation AI-enabled microscopy ecosystems, including:

✔ Agentic control of instruments – connecting LLM-driven agents to microscopes via APIs and orchestration layers.

✔ Reward-driven experimental design – formalizing objectives and enabling autonomous decision-making beyond heuristic workflows.

✔ Cloudification and federation – transforming microscopes into distributed, connected platforms operating across sites.

✔ From data to action – closing the loop between acquisition, analysis, and decision-making in real time.

Participants will work with realistic microscopy datasets and digital twins, enabling development and testing of ML workflows under conditions that reflect actual experimental constraints. The emphasis is on transitioning from human-controlled experiments to AI-augmented and ultimately agent-driven systems, where decision-making happens at the time scale of the instrument.

Hands-on sessions and hackathon-style modules will ensure that participants engage directly with:

  • API-level control and orchestration of instruments
  • Integration of ML models into live experimental workflows
  • Stochastic optimization and decision-making under uncertainty
  • Coupling between data streams, models, and actions

Whether your interest is in microscopy data analysis, automated control, or building the next generation of cloud-connected, AI-driven scientific instruments, this school provides a concrete pathway from concepts to implementation.