MetaCell

Filed from Cambridge, Massachusetts  •  Company Desk

Here is a fact that sounds like a setup to a joke but is not: before MetaCell was a software company selling to Johnson & Johnson and Pfizer, its founders were trying to build a worm. Not a real worm - a digital one, neuron by neuron, all 302 of them, in a project called OpenWorm. The nematode C. elegans is about as simple as a nervous system gets, which is exactly why people thought it might be simulatable. It has, so far, resisted. And that resistance turns out to be the whole point.

Because when you spend years trying to model a living thing from its raw data - microscopy images, electrophysiology traces, connectomes - you learn, in a very expensive and personal way, why life-science data is such a mess to work with. It comes in a dozen formats. It doesn't line up. The tools to visualize it barely exist. Most people would file that under "reasons this is hard." MetaCell's founders filed it under "business."

So the pitch is straightforward, and the company doesn't dress it up much: MetaCell is a life-science software company staffed by scientists and software engineers who take the kind of data that makes researchers groan - MRI, EEG, calcium imaging, biologically accurate simulations - and turn it into software that people can actually use, share, and analyze. They describe themselves, without much hedging, as "world leaders in software for life science," and then, tellingly, list their clients instead of piling on adjectives.

The unusual hiring rule

The thing that makes MetaCell genuinely different is a staffing decision most companies can't afford to make. The people building the software tend to have both a PhD in a life science and a career in software engineering. Not a scientist paired with an engineer who translate for each other across a table - the same person, holding both. CEO Stephen Larson is the template: a bachelor's and a Master of Engineering in computer science from MIT, then a PhD in neuroscience from UC San Diego. His co-founders, Giovanni Idili and Matteo Cantarelli, come from software architecture and systems engineering.

Why does this matter? Because in most technical consulting, the expensive failure mode is translation loss - the scientist explains the data, the engineer builds something almost right, and three sprints disappear into the gap between them. Remove the gap and you remove the failure. It's a costly way to hire. It's also, apparently, why a 28-person shop can credibly serve the Allen Institute and the Chan Zuckerberg Initiative at the same time.

The open-source flywheel

MetaCell also did the thing that sounds like charity but functions like marketing: it built an open-source platform called Geppetto, a modular system for visualizing neuroscience models and managing simulations, and gave it away. The academic paper on it landed in the Philosophical Transactions of the Royal Society B. And then the work showed up: UCL contracted MetaCell to build features of Open Source Brain, EMBL-EBI for the Virtual Fly Brain, the State University of New York for neuroscience UIs. The free thing advertised the paid thing. This is not an accident anyone should be surprised by, but it's rare to see it executed so cleanly.

The productized side of the business is Cloud Workspaces - computational notebooks on demand, with compute that scales to whatever you're running, so a researcher spends the afternoon on the science instead of on environment setup. There's Cloud Hosting for life-science and healthcare workloads, NeuroGlass for wrangling microscopy data, and Cloud Harness for building the Kubernetes plumbing underneath. Underneath all of it is the least glamorous product in the catalog and possibly the most valuable one: data standardization. Nobody dreams about it. It's also the difference between finding an answer and never finding it.

The moonshot still runs

What's quietly interesting about MetaCell is that the worm never got shelved. The founders run both the company and the nonprofit OpenWorm Foundation, and the arrangement is symbiotic in a way more mission-driven founders should study: the company funds the science, the science feeds the company's expertise, and neither has to pretend the other doesn't exist. You don't have to choose between the moonshot and making payroll. Structured right, one pays for the other.

The virtual worm, to be clear, is still not finished. Simulating even the simplest nervous system remains one of biology's genuinely open problems, and MetaCell's founders have been at it for more than a decade. The tidy narrative would call that a failure. The less tidy and more accurate version is that a bet too big to finish still produced real tools, a real client roster, and a real company. Ambition compounds even when it doesn't complete - and in MetaCell's case, it compounded into software that the people studying the human brain now depend on.