founder-built
Medicine, computation, and biology in one founding team.
A physician-computer scientist and a computational biologist met at Carnegie Mellon and built Galen's virtual-cell core together.
The virtual cell company.
Most of the field models snapshots of biology.Galen models its trajectory over time.
Galen’s virtual cell predicts how real cells change over time in response to genetic, chemical, or environmental interventions.
Review the evidence, real-T-cell proof plan, and milestones with the founders.
the pre-seed opportunity
Apply our virtual cell to cell therapy R&D.
01 | wedge
T-cell durability
Engineered T cells can lose function under repeated tumor pressure. Only a fraction of candidate changes can receive full real-cell testing.
02 | product
A decision-ready program
Cell-therapy R&D leaders use Galen to decide which candidate changes deserve assays. Programs return a ranked shortlist, uncertainty, a locked scoring plan, and an evidence certificate for $150K to $750K.
03 | proof today
Three capabilities built
Cell-state recognition and calibrated uncertainty are benchmarked on held-out real human-cell states; intervention dynamics in controlled simulation. All three form the computational core.
04 | proof next
Real T-cell scoring
Before a real T-cell screen, Galen locks the ranking, comparator, and scoring rule. Success means outperforming the agreed reference at identifying interventions that preserve durable function under repeated challenge.
05 | round
$2M | 18 months
Raising $2M over 18 months to sign the first biopharma program, score the real-T-cell comparison, and repeat. Meeting investors through August 2026.
why Galen can win
A focused decision product with an earned path to platform scale.
why now
Measurement, method, and budget have converged.
Time-aware single-cell and perturbation measurements now support model-guided experiments, while cell-therapy teams already fund persistence and construct-optimization decisions.
defensibility
Each scored cycle improves the next.
Galen owns the prediction, ranking, evidence contract, and scored decision product. Returned measurements sharpen the next queue under agreed data frameworks.
expansion
One program expands across an asset.
A $150K to $750K entry program can expand into $1M to $3M in recurring annual cycles per therapeutic asset, then into multi-program relationships after transfer is demonstrated.
the technical thesis
Galen models the cell as a system under intervention.
Most of the field builds better descriptions of a cell at a moment in time. Galen models what a cell does when you change it. Those are different mathematical objects.
Atlases and cell foundation models are trained on observation. They learn what varies with what, which is correlation. But correlation does not imply causation. Forecasting an intervention that has never been tried is a causal question, and no amount of observational data answers it.
what Galen does instead
A virtual cell usually predicts a snapshot: the state a cell lands in after a change, learned from changes already observed. Galen predicts the trajectory. It learns the rules that carry a starting state forward, then treats an intervention as an operation on those rules rather than as another input. The prediction continues after the intervention stops. That is where a durable effect separates from a temporary one.
what the field builds
Description
Organize and represent the biology that has already been measured.
what therapeutics need
Intervention
Forecast the outcome of a change nobody has run yet, and rank it against the alternatives.
Because the intervention is part of the model, Galen can rank changes it has never seen. Each prediction is held to a locked scoring rule, and the returned measurement sharpens the next.
why virtual cells
A world model of human biology.
Cells are where disease takes hold and medicine takes effect. A virtual cell is software that predicts how a real one behaves. Give it a starting state and a proposed intervention, such as a drug or a gene edit, and it forecasts how that cell changes over time.
Virtual cells will let scientists test more ideas in software, spend scarce experiments on the strongest candidates, and reach a working therapy sooner.
01 | measure
Establish the starting point
Define the cell's starting state and the candidate interventions.
02 | forecast
Model what changes
Predict how the cell changes under repeated biological pressure.
03 | rank
Focus the search
Prioritize candidates and report uncertainty.
04 | test and learn
Return to living cells
Test the locked shortlist in the lab; returned measurements score the result and improve the next cycle.
A model of life must remain answerable to actual wet lab biology. Real cell biology experiments score the virtual cell predictions and determine the next tests.
The broader virtual-cell platform will be built one valuable, measurable, repeatable decision at a time. Galen begins with T-cells, specifically tackling the challenge of durability and exhaustion in engineered cell therapies.
the founders
Built by the founders. Ready for biopharma programs.
A physician-computer scientist and a computational biologist designed and built Galen's computational core together.
Co-founder and CEO
Logan Nye, MD
Logan is a physician and former Harvard Medical School clinical-AI researcher trained in computer science at Carnegie Mellon. He leads clinical judgment, causal evaluation, and product direction.
Co-founder and CTO
Kushagra Agarwal
Kushagra is a computational biologist trained at IIIT Hyderabad and Carnegie Mellon. He leads model architecture, biological data systems, and scientific software.
the opening
Biology has reached the intervention era.
Measurement, method, and budget have arrived at the same time. Galen enters that opening with its computational core already built.
measurement
Measurement now resolves response over time.
Perturbation and single-cell assays now provide time-aware inputs for model-guided experiments.
technical opening
Biology is moving from description to intervention.
Galen is that layer. It predicts change, quantifies uncertainty, and improves the next decision.
buyer readiness
Cell-therapy teams already fund these decisions.
Persistence, construct optimization, and experimental prioritization already sit inside therapeutic R&D budgets.
Cell maps organize observed biology. Laboratories measure what happened. Galen ranks, locks, and scores the next experimental decision.
the value at stake
$343 billion a year
Global R&D spend that better biological decisions can redirect
About 90% of drugs entering the clinic fail, and preclinical attrition from hit to candidate exceeds 90%. The average cost per approved drug reaches roughly $2.5 billion because every failure is absorbed by the one candidate that works.
Galen ranks those candidates before the spend commits.
the beachhead
Make engineered T cell therapies last.
Scientists engineer immune cells to recognize cancer, manufacture them as a therapy, and deliver them to the patient. Under repeated tumor pressure, those cells can become exhausted and lose their cancer-fighting function.
Durability is an urgent and expensive design problem for immuno-oncology. Every weak experiment consumes scarce cells, time, and assay budget. Galen concentrates that investment on the candidates with the strongest predicted durability.
Teams prioritize with prior assays, scientific judgment, internal analysis, and partial screens. Galen ranks the candidates before scarce experimental capacity is spent.
why this starting point
A valuable decision. A measurable endpoint. A compounding product.
T-cell durability combines an existing development budget, a repeated-challenge endpoint, and a search space that rewards predictive prioritization.
01 | valuable
Teams already fund persistence and construct-optimization decisions.
Galen enters an existing therapeutic-development budget.
02 | measurable
Repeated challenge reveals whether engineered function endures.
Durable function creates a clear product endpoint.
03 | strategic
Each cycle adds intervention evidence to the product.
Each program leaves the model stronger for the next one.
the product
A T-cell durability program, end to end.
A production decision system for therapeutic R&D teams, combining predictive software with a locked evidence workflow.
the deliverable
Galen delivers ranked candidates, confidence estimates, a locked screen and scoring plan, an evidence certificate, and the next measurement queue.
01
Program input
T-cell context, genes and receptor designs, combinations, culture conditions, prior measurements, and assay budget.
02
Galen prediction
Predicted repeated-challenge trajectories, uncertainty, and a ranked shortlist.
03
Locked test
Candidates, comparator, measurements, and scoring rule fixed before results exist.
04
Decision output
Returned measurements produce a scored result, evidence certificate, and next queue.
evidence certificate
An auditable record of what Galen predicted, what was tested, how the result was scored, and what should be measured next.
business model
The program fee covers model setup, ranking, uncertainty, prospective scoring, and returned-result analysis. Real-cell execution is scoped with the program, so Galen carries software economics and no laboratory capital cost.
Initial program
$150K to $750K
T-cell durability program
One candidate library, ranked and scored across an agreed number of measurement cycles.
Program scope sets the range: candidate-library size and the number of ranking-and-measurement cycles.
Asset expansion
$1M to $3M annually
Recurring virtual-cell cycles per therapeutic asset
The same intervention workflow guides successive decisions across one cell-therapy program.
Platform expansion
$3M to $10M annually
Multi-program virtual-cell partnership
Platform access extends the workflow across programs, teams, and intervention libraries.
Partnered programs
Milestone economics
Development partnerships on ranked therapies
Galen participates in the value of the assets its scored rankings shape.
current proof
A working virtual-cell core, benchmarked end to end.
Ranking an intervention rests on three capabilities: reading where a cell starts, forecasting where a change takes it, and knowing when that forecast can be trusted. Any one alone produces a description; together they produce a decision. Galen runs all three as one system. Each is benchmarked below.
01 | Real human cell state recognition
Know the cell's starting state.
98.9%
On held-out real human cells, Galen accurately recognized the cell state it was given.
A virtual cell must begin from the right biological state before it can forecast what happens next.
Known-lineage recognition on held-out cells; about 72% for the matched linear reference
02 | Controlled simulated benchmark
Distinguish lasting change from a temporary response.
0.03 versus 0.82
With intervention response withheld, Galen finishes 0.03 from the hidden target versus 0.82 for the matched simpler mechanism. Lower is better.
In a controlled simulation, Galen preserved the difference between a durable state change and a temporary push, central to predicting whether engineered function will persist.
Lower is better; 0.82 distance for the matched simpler mechanism
03 | Real human cell state calibration
Know when to trust a prediction.
90% target met
Galen's uncertainty intervals achieve their stated 90% coverage target on held-out human-cell states.
Calibrated uncertainty separates confident rankings from the candidates that need more measurement.
90% interval-coverage target achieved on held-out human-cell states
04 | pre-seed milestone
Rank real T-cell interventions under repeated challenge.
Locked real-cell screen
Galen commits the ranking, comparator, and scoring rule before the screen runs, then scores the result against returned measurements.
Agreed reference defined before results exist
The core components are built, benchmarked, and integrated. The pre-seed puts them to work inside biopharma T-cell programs.
the market
From a first program to partnered economics.
Virtual-cell leadership will belong to the system researchers trust to choose what happens next. Galen's revenue architecture climbs from focused entry programs to development partnerships in the therapies it shapes.
01 / deploy
pre-seedEngineered T-cell durability
what it proves
Score a prospectively locked ranking for durable function through repeated challenge.
what it unlocks
First $25M to $100M of ARR
100 to 200 qualified accounts × $250K to $500K annually
02 / repeat
pre-seedRecurring cycles within one asset
what it proves
Reuse the workflow across successive intervention decisions in a therapeutic program.
what it unlocks
That ARR becomes recurring
03 / transfer
A second biological context
what it proves
Demonstrate that the core workflow transfers without rebuilding the intervention engine.
what it unlocks
Proves the platform thesis
04 / expand
Multi-program immune-cell relationships
what it proves
Extend the shared engine across programs, teams, and intervention libraries.
what it unlocks
$318M to $636M annually
424 developers × $750K to $1.5M annually
05 / platform
A common virtual-cell runtime
what it proves
Build a shared foundation for intervention applications across biological contexts.
what it unlocks
$1B to $3B annually
Eligible programs × decision cycles × annual intervention, assay, optimization, and process-development spend
06 / partner
Development partnerships on the therapies Galen ranks
what it proves
Run scored rankings inside partner programs and share in the assets they shape.
what it unlocks
Milestone economics per asset
114 AI-discovery partnerships signed in 2025 for $43.4B in potential payments; single platform deals now clear $2.5B
Source: AI discovery partnership volume, 2025Source: Single-deal benchmark, 2026
Each layer is a distinct economic surface with its own unlock. Galen's monetization begins with qualified T-cell accounts and current program pricing, then climbs as scored rankings earn a share of the assets they shape.
The wedge creates a reachable first business. Repeatability, transfer, and partnered economics create the asymmetric upside across software, therapeutic development, and the capital directed toward biological R&D.
the pre-seed
Raising $2M. Three gates in 18 months.
Three gates retire the three risks in order: will anyone pay, does it work, does it compound. Galen advances technical, experimental, and commercial work in parallel, moving as each gate clears.
SIGN
Work
Integrate wet lab execution, secure the first biopharma program, and lock the comparator, scoring plan, and measurement-return contract before any work begins.
Evidence unlocked
A signed program with the reference and the scoring rule agreed in advance.
Customer + company value
A named partner, contracted program revenue, and a candidate library to rank.
SCORE
Work
Rank the program's candidate library, run the locked screen through repeated challenge, and score the returned measurements against the agreed reference.
Evidence unlocked
A locked real-T-cell ranking scored against the agreed reference on durable function.
Customer + company value
A decision-ready result, evidence certificate, and next experimental queue.
REPEAT
Work
Use the returned measurements to update the model, rank the next queue, and complete a second cycle inside the same program.
Evidence unlocked
A second scored cycle that improves on the first.
Customer + company value
A repeatable customer product packaged for recurring use within a therapeutic asset.
capital advances
- Experimental T-cell biology and scientific-machine-learning talent.
- Virtual-cell model, data system, and compute.
- Wet lab integration and real-cell experimental capacity.
- Two ranking-and-measurement cycles inside the first program.
- A repeatable virtual-cell customer product.
the outcome
The pre-seed turns a signed program into a scored result and a second cycle that improves on it. Transfer into a second biological context is the seed milestone.
founder discussion
Discuss Galen's pre-seed with the founders
Request a direct conversation to explore the company, review private round materials, and discuss participating in the round.
Raising $2M. Meeting investors through August 2026.