At pre-seed, sourcing is a search problem. The companies that will define healthcare are already alive, as a result that holds, a patent no one has read, a clinician’s hunch, scattered across labs and journals long before any of it looks fundable. The edge is seeing them before anyone else does, and being literate enough to know what you are looking at.
To widen what we can see, we lean on Value Alpha, a research and machine-learning effort built to surface early signal systematically. Its core algorithm was developed in collaboration with Columbia Business School and Columbia’s engineering and machine-learning community, and is led by our partner Tomasz Felpel, who is also a partner at Value Alpha.
Models widen the funnel. Conviction and literacy still decide.
What it does
Value Alpha reads the boundary between research and company formation, the publications, patents, technology-transfer activity, and clinical signals that precede a fundable company, and helps us rank where the first institutional dollar is most likely to change an outcome. It is a way to be early at scale, not a substitute for judgment.
Why Columbia
Sonnerie was born from the Columbia ecosystem, and so was this work. Building the algorithm with Columbia Business School and the engineering school pairs domain understanding with machine-learning rigor, the same combination we look for in founders: people close enough to the science to judge it, and close enough to the market to build it.
The output is never a decision; it is a better starting point. Every name still meets an operator who has built, exited, and navigated regulation in healthcare. The machine helps us listen across more of the field. The judgment, and the conviction to write the first check, stays human. From signal to scale.