The data platform for research.
Every figure in a paper traces back through its tables and cells to the raw data.
- tablecohorts.effects
- cellfit_model
- cellclean_visits
- fetchs3://lab-raw/visits/
- cellsource, inputs, environment, author
- artifactevery result, cached by what made it
- tableParquet under Iceberg, in your organization's own prefix
- publicationpermanent page and DOI
- paperfigures, snapshots and their DOIs together
From the first notebook to the published paper.
Notebooks that remember
Hosted Strata notebooks where every result is cached by what made it, so nothing expensive runs twice.
Compute when a cell needs it
Send one cell to a CPU or GPU worker, or to the lab's own machine.
A DOI for the figure
Publish a figure with a permanent page of the code behind it and a DataCite DOI for the paper.
Double-blind review
Share a frozen snapshot with names withheld. Reviewers read it in the browser.
Restricted data stays put
Nothing derived from restricted data can be published or shared outside the organization.
Agents with a paper trail
Agents work with budgets and approval gates, and every cell records whether a person or an agent wrote it.
Your tables, readable by every engine.
After the first release, promoted tables will be written as Parquet under Apache Iceberg in your own storage, so DuckDB, Spark, Trino, Snowflake and Databricks can read them. Adding Amber needs no migration, and leaving it takes your data with you.
Research groups first, then departments.
Amber starts with groups of one to fifteen people with a paper in progress and data they are careful with, then grows into what a department or core facility needs. Publishing is free on every plan.
Publishing a paper in the next year?
We are looking for a few research groups to work with on real papers.