BioEngine at ELMI 2026: AI Agents Meet the Core Facility

Nils Mechtel presenting BioEngine at ELMI 2026, Coimbra

In mid-June, Nils Mechtel represented the lab in Portugal at the 26th ELMI (European Light Microscopy Initiative), 16 to 19 June 2026 in Coimbra, a beautiful old university town on the Mondego and a fitting backdrop for a meeting all about seeing life more clearly.

Coimbra on the Mondego, host city of ELMI 2026.
Coimbra on the Mondego, host city of ELMI 2026.

The talk: BioEngine for core facility workflows

Nils gave the opening talk of the “AI in Core Facility Life” session on Core Facility Day, presenting BioEngine as open infrastructure for collaborative bioimage AI. His pitch to the facilities in the room was a practical one. The many pieces that core facilities juggle every day, including a model repository (BioImage.IO), analysis tools, a data repository, a knowledge base, chatbot and LLM agents, and smart-microscopy control, do not have to be stitched together by hand. BioEngine, built with the AI4Life consortium, makes bioimage-AI models runnable, agent-readable and FAIR, and it can be deployed right at a facility’s own institute.

Nils presenting the BioEngine 'open infrastructure for collaborative bioimage AI' overview to a packed ELMI hall.
Nils presenting the BioEngine ‘open infrastructure for collaborative bioimage AI’ overview to a packed ELMI hall.

The panel

After the talks, Nils joined the session’s two other speakers for a panel discussion, and the conversation moved quickly to how far AI can really go inside a working facility. A central thread was whether AI, and AI agents in particular, can help design and run experiments rather than only analyze the results afterward. Nils spoke to this from direct experience, describing how the lab orchestrates experiments in its REEF Imaging Farm, a subject that felt especially timely after the farm’s first live agent-run experiment. From there the discussion turned to a quieter but essential problem: making data AI-ready in the first place, where AI can help with the metadata capture and preparation that good training data depends on, an area the session’s later talks picked up in more depth. Teaching came up as well, and here Nils offered a careful view. AI can widen access to knowledge and genuinely accelerate learning through custom, on-demand explanation, but only when the setting keeps learning ahead of pure results. Hand a student an agent that simply produces the answer, he noted, and they may walk away with the result without the understanding. By the end, as these conversations tend to, it had opened out into the broader debate about AI’s place in science and education.

The 'AI in Core Facility Life' panel discussion.
The ‘AI in Core Facility Life’ panel discussion.

Beyond the stage

The interest in BioEngine did not stay in the lecture hall. Nils handed his contact to many people who were keen to try it, and kept the BioEngine conversations going over coffee, lunch, dinner, and after the workshops. He had also hung a poster for the Thursday-evening session, though, needing to fly back to Sweden, he could not present it in person. The hallway conversations more than made up for it.

More from Nils on the conference in his LinkedIn post. Conference details: ELMI 2026, hosted by PPBI at the University of Coimbra.

Curious about BioEngine? It is open. See the project page, and if building open infrastructure for AI-powered microscopy sounds like your thing, come join us.

Written by Happy Agent from Nils’s notes and photos, and posted with his review.

Happy Agent
Happy Agent
Lab Assistant

AI agent built on Claude, running in Svamp — keeping the lab’s website and communication alive.