Transfyr, a physical AI platform for real-world science, has launched with $25 million in seed funding led by General Catalyst, with participation from Lux Capital, Breakout Ventures, Factory, Neo, SV Angel, MVP Ventures, Underscore VC, Lyda Hill, and several angel investors. Founded by industry veterans Anna Marie Wagner and Dr. Renee Wegrzyn, the Cambridge, MA-based company aims to build a lossless understanding of scientific execution to accelerate translation, workforce development, and physical AI applications.
The Problem: Science's Missing Layer
Everything from scientific reproducibility to robotics is bottlenecked by an incomplete understanding of the contextual dependencies and physical realities of hands-on lab work. Transfyr is building the bridge between the physical and digital divide in scientific laboratories, capturing bench science and converting it into machine-readable data for closed-loop systems powered by AI and automation.
AI can ingest scientific data and published articles, but it can't learn from what's missing from the scientific record — hard-won lessons from failure, the tricks to overcome finicky protocol steps, and tacit knowledge that's never written down. Today, successful lab-to-lab handoffs still rely on apprenticeship, often taking months of training, troubleshooting, and bespoke tech transfer, and many still fail.
"Science is missing a critical layer of infrastructure that's necessary for efficient reproducibility, translation, scaling, and automation."— Anna Marie Wagner, Co-Founder and CEO, Transfyr
How the Platform Works
Transfyr deploys integrated sensor systems and multimodal models trained on real-world scientific execution to passively capture and interpret what's missing from the scientific record. The platform learns each customer's context to build a reliable record of operator actions and intent, environmental context, equipment telemetry, and supply chain dynamics.
This metadata can surface sources of process variability, enable root cause analysis, optimize protocols, create training and tech transfer SOPs, and build robotic-level instructions — all powered by active reinforcement learning loops. Headquartered at The Engine in Cambridge, MA, Transfyr also operates an in-house wet lab where it generates foundational training data, tests its sensor stack in real experimental workflows, and runs evaluations for top frontier labs.
"Breakthroughs mean nothing if they stay trapped in a single lab or depend on unwritten tacit knowledge to succeed. The real bottleneck to revolutionary science isn't a lack of big ideas, it's the massive friction of translating those ideas into reliable, scalable reality with impact."— Renee Wegrzyn, PhD, Co-Founder and Chief Innovation Officer, Transfyr