Profile / Builder of Machines That Read Molecules
Jack Geremia
He trained machines to steer quantum systems at Caltech. Now he is teaching them to read the raw signal of a molecule.
He trained machines to steer quantum systems at Caltech. Now he is teaching them to read the raw signal of a molecule.
Inside a converted building on Somerville Avenue, Jack Geremia runs a company betting on a single, strange premise: that a machine can look at the raw squiggle of a mass spectrometer and understand what it is seeing. No library lookups, no painstaking manual curation - just a model, trained on billions of spectra, reading molecules in minutes.
That product is the Large Spectral Model, and Matterworks calls it an LSM on purpose. The analogy to the large language model is the whole point. Where an LLM learns the grammar of language from oceans of text, the LSM learns the grammar of molecules from oceans of spectral signal. Geremia, who became CEO after co-founding the company in 2019 as its advising CTO, is the person tasked with turning that bet into a business.
His pitch is blunt about the problem it solves. Mass spectrometry is one of the most powerful instruments in the life sciences and one of the most stubbornly inaccessible - slow, expensive, and gated behind specialist expertise. Matterworks wants to make it as routine for a working biologist as single-cell sequencing became. The phrase Geremia keeps returning to is the unsustainable cost of life science R&D.
In 2024 the company shipped Pyxis, billed as the first generative AI for complete untargeted molecular annotation from raw mass spec. In June 2025 it raised a Series A. Somewhere between those two events, the advising CTO became the chief executive.
“ accelerate machine intelligence innovations that tackle the unsustainable cost of life science R&D - Jack Geremia, on the 2025 Series A
The throughline is not a field. It is a habit: take a system nobody can read, and build the thing that reads it.
A postdoctoral fellowship in the Division of Engineering & Applied Science - studying how to steer physical systems. An odd foundation for a biotech founder, and a telling one.
His first build in the microbiome space. The company was later acquired by DSM.
Running microbiome discovery at scale inside one of the world's largest nutrition companies.
Back to founding - leading technology development at the US venture.
Driving research and development at the Boston-area microbiome company.
With Mimoun Cadosch Delmar. Geremia starts as advising CTO - the science conscience, not the CEO.
The company launches Pyxis and Geremia delivers a lightning talk: When will phenotype prediction be reality?
Lewis & Clark Partners and OMX Ventures lead, with Pillar VC, Germin8, Intermountain, and Tarsadia joining. Carolyn Fritz joins the board.
Self-supervised, trained on billions of spectra, built to capture chemical and biological relationships directly from raw signal - in minutes, not weeks.
Unstructured mass spectrometry data - the messy, high-dimensional output most pipelines need experts and reference libraries to interpret.
The LSM, pre-trained on billions of spectra, interprets the signal the way a language model interprets a sentence - learning the underlying grammar of molecules.
Identification and absolute quantitation of metabolites, ready to predict complex biological outcomes - delivered as a model-as-a-service.
He joined his own company as advising CTO and stepped up to chief executive years on - the reverse of the usual founder-to-figurehead drift.
Quantum control and dynamical systems are not where biotech CEOs usually train. That control-theory instinct shows up in a company built on steering data.
Microbiome discovery for animals, R&D for human therapeutics, and now AI for biology writ large - he keeps founding across the whole stack of life sciences.
A prolific inventor's paper trail - the kind of body of work that takes decades and a refusal to sit still.
The flagship product borrows its name from a constellation named for the navigator's compass - a quiet tell about how this team thinks about finding its way.
His investors compare the ambition to single-cell sequencing - taking a specialist instrument and handing it to every biologist.
“Matterworks is leveraging AI to unlock a new data layer for biology by making mass spectrometry accessible to all biologists. It's a shift as transformative as single-cell sequencing and we are thrilled to support this exceptional team.”
“Matterworks has repeatedly demonstrated its AI platform provides a step change reduction in the cost and throughput of mass spec quantitation. We are excited to partner with this team in accelerating commercialization of their transformational technology.”
His postdoc was literally in control and dynamical systems - he studied how to steer physical systems before steering molecular-data companies.
Three degrees, three coasts: UC Davis, Delaware, Princeton - plus the Caltech fellowship.
He answers to both Jack and his initials J.M. Geremia in professional life.
Matterworks names products after navigation - Pyxis is the constellation of the compass.