Transfyr has launched with $25M in seed funding to build what it calls an observability layer for science: sensors and multimodal AI models that capture what actually happens at the lab bench.
General Catalyst anchored the seed, with Lux Capital, Breakout Ventures, Factory, Neo, SV Angel and Underscore VC among those joining.
The Cambridge, Massachusetts startup was founded by Anna Marie Wagner, former head of AI and corporate development at Ginkgo Bioworks, and Renee Wegrzyn, the founding director of ARPA-H. Their thesis is blunt: AI models can ingest papers, protocols and databases, yet the formal scientific record omits the details that determine whether an experiment can be reproduced, such as failed attempts, equipment tweaks, environmental conditions and operator workarounds.
“Science is missing a critical layer of infrastructure that’s necessary for efficient reproducibility, translation, scaling and automation,” said Wagner, who serves as CEO.
By translating bench activity into machine-readable data, Transfyr aims to feed a new generation of autonomous lab systems. It arrives as AI companies push beyond text and software into physical environments, where the gap between written method and real practice is widest.
