Learn without labels
Find the latent patterns in your images before you know what to call them.
Representation learning for cell imaging
Your microscopy images
already know more.
Percolatech turns raw cell images into a biological map your team can search, compare, and build on — without hand-labeling every phenotype.
10×faster phenotype
discovery
discovery
0labels needed
to start
to start
∞questions from
the same image set
the same image set
LIVE ASSAY / 038PHENOTYPIC SPACE
cluster 04
rare state
mitotic
RAW PIXELS → BIOLOGICAL SIGNAL
The shift
Stop treating images
as evidence. Start using
them as data.
Cell imaging holds the record of what your biology is doing. But today, most teams only inspect a sliver of it — after weeks of manual annotation and brittle analysis.
Percolatech learns directly from your image collections. The result is a high-dimensional representation that makes subtle cell states measurable, comparable, and reusable across every experiment.
Explore the platform ↘Measure what changes
Quantify treatment response, morphology, and state transitions in one shared space.
Reuse every experiment
Build a continuously more valuable representation of your biology, not another silo.
From pixels to possibility
One simple loop.
Compounding insight.
Built for the reality of modern cell biology: messy images, evolving questions, and science that can’t wait for perfect labels.
Connect your images
Upload a plate, point to your cloud bucket, or integrate your imaging pipeline.
OME-TIFF · PNG · JPG · HCSPercolate learns
Our models build a robust representation of morphology, structure, and cell state.
SELF-SUPERVISED · MULTI-SCALEExplore new biology
Surface novel phenotypes, compare conditions, and export features to your stack.
SEARCH · CLUSTER · PREDICTBuilt for discovery teams
A better lens for the questions that matter.
PHENOTYPIC SCREENING
See hits
you weren’t
looking for.
Find and rank subtle response patterns across large screens.01PERTURBATION BIOLOGY
Map a cell’s
response
to change.
Compare genetic and chemical perturbations in a common space.02MODEL DEVELOPMENT
Build on
biology that
gets smarter.
Turn your image archive into a living foundation model.03A representation you own
Your images.
Your biology.
Your advantage.
Percolatech is designed to become a quiet, powerful layer in your research stack — giving your team a durable way to learn from visual biology.
- Private-by-design workspaces
- Exportable embeddings and features
- Human-readable model reports
ASSAY SPACE / CONTROLS + TREATMENTS
untreated
compound A
novel state
Not another black box
Every run leaves your team with something useful.
Searchable phenotype space
Browse any experiment by visual similarity, surface outliers, and turn the clusters you find into a shared biological vocabulary.
Exportable model outputs
Take embeddings, per-cell features, and condition-level summaries into your notebooks, LIMS, or downstream machine-learning workflows.
Decision-ready readouts
Trace every signal back to the image evidence so scientists can validate, prioritize, and communicate what the model has found.
Straightforward plans
Scale insight,
not overhead.
Every plan includes our self-supervised cell representation engine, a private workspace, and scientist-led onboarding. Start with the image volume you have today.
EXPLORE
For a focused assay or early program.
$1,250/ month
billed annually
Start with Explore- Up to 250,000 images / month
- One private workspace
- Core embedding + search tools
- CSV and Parquet exports
- 48-hour support response
MOST CHOSEN
DISCOVER
For teams running repeatable imaging programs.
$4,250/ month
billed annually
Choose Discover- Up to 2 million images / month
- Five named collaborators
- Plate and condition comparisons
- API + notebook starter kit
- Priority scientific support
FOUNDATION
For organization-wide visual biology.
Custom
annual platform agreement
Talk to our team- Unlimited workspaces and seats
- Custom storage and image volume
- Private cloud or VPC deployment
- SSO, audit logs, admin controls
- Dedicated scientific partnership
Need a proof-of-value first? We run a 30-day pilot on one of your real assays. Plan a pilot.
Built for sensitive science
Your data stays in your orbit.
Percolatech was designed for teams working with valuable, pre-publication biology. You decide who can access a workspace, what gets exported, and where your image data lives.
Talk to security ↗Private by default
Your images and derived representations are isolated from every other customer workspace.
Enterprise controls
Role-based access, audit trails, and SSO are available on Foundation.
Deployment flexibility
Use our managed cloud or keep data inside your approved cloud environment.
Questions, answered
Before we work
with your images.
What types of images can Percolatech learn from?
We support common microscopy image formats including OME-TIFF, TIFF, PNG, and JPEG. During onboarding, we help map channels and plate metadata so your assay is ready to learn from.
Do we need labels to start?
No. Percolatech learns directly from image structure, morphology, and variation. Labels and assay metadata can be added later to test hypotheses or train targeted classifiers.
How quickly can a team be live?
A focused pilot can begin as soon as we receive a representative assay and its metadata. Most teams have a first searchable representation within days, not months.
Can we use our own storage and cloud environment?
Yes. Foundation includes private-cloud and VPC deployment options for organizations with specific data residency or security requirements.
The frontier is already in your images
Let’s learn
what it’s trying to tell you.
Bring us one assay and one question. We’ll show you what a learnable representation can uncover.
Plan your pilot ↗