Meet Gerrit Renner

Gerrit develops data-processing and chemometric approaches that preserve the uncertainty, quality, and provenance of analytical measurements. His research aims to make data trustworthy throughout the measurement chain – from the moment a signal is recorded to the scientific conclusions ultimately drawn from it.

“What excites me most is treating data processing as an integral part of the analytical measurement rather than as something that happens afterward,” he says. Steps such as centroiding, feature detection, integration, and modeling all introduce assumptions and uncertainty. Gerrit’s group is developing ways to make these transformations transparent and quantitative, with uncertainty estimates and quality measures that travel through the workflow. “If we want machines to make meaningful use of analytical data, they need more than numbers: they need information about where those numbers came from and how trustworthy they are.”

AI may capture the attention, but Gerrit believes machine-actionable analytical data will bring the more fundamental change. “Trustworthy AI in analytical science will ultimately depend less on ever-larger models and more on giving those models better-structured, traceable, and scientifically meaningful data,” he says. That will require results to retain their metadata, provenance, uncertainty, and semantic meaning rather than leaving essential context buried in proprietary formats, software settings, or laboratory notebooks.

“FAIR data should not be the final step of a research project; it should be a property of the entire analytical workflow,” Gerrit argues. Trying to make results findable, accessible, interoperable, and reusable only after processing risks losing critical context. Designing FAIR principles into acquisition and analysis from the outset would improve reproducibility and give AI systems a firmer scientific foundation.

Gerrit’s career took shape when he decided not to choose between analytical chemistry and data science. “The measurement does not end at the detector,” he says. “Data structures, algorithms, statistics, and scientific interpretation are all part of the same measurement chain.”

Knitting has reinforced a similar lesson: “Complex knitting patterns are built from a limited number of simple operations, but small mistakes can propagate and become surprisingly difficult to correct later. It teaches patience, pattern recognition, and the importance of understanding structure rather than simply following instructions. I've also learned a research lesson: sometimes the most efficient way forward is to go back, find where something went wrong, and reconstruct it properly rather than patching the final result.”