Become a design partner
Running a large, multi-team object-storage estate? Work with us on real findings from your buckets and help shape how Varve identifies datasets, owners, cost, and what can move, expire, or stay.
Apply as a design partner
Outcrop is live
Give object storage a system of record.
Varve reads object metadata and activity signals to identify datasets in your storage: who owns them, how they're used, what they cost, and what can safely change.
Read-only. Metadata, never contents.
In active development and private preview with selected design partners. Not generally available yet.
s3://data-lake
events/year=2025/month=06/part-0007.parquetevents/year=2025/month=07/part-0031.parquetevents/year=2025/month=08/part-0094.parquet + 1.2B more objects Varve discovery engine
events dataset
What it is
After a few years, most teams can list the objects in storage but not explain the datasets.
S3 Inventory lists objects. Access logs show activity. Neither tells you which objects form a dataset, when that dataset landed, who is accountable for it, or what might break if you move it.
The system has objects, but the organization thinks in datasets.
Varve turns those signals into an owned record of the datasets in your storage: something teams can inspect, govern, and safely change.
Proof
See the engine, running in public.
Outcrop is Varve's public catalog: open object-storage datasets exposed as dated, readable records. It shows the discovery engine on data you can verify yourself.
Outcrop and Varve stay distinct: Outcrop is the public catalog; Varve is the commercial system of record for private storage estates.
- Common Crawl
- Sentinel-2
- NEXRAD
- OpenStreetMap
- 1000 Genomes
s3://sentinel-s2-l1c public
verifiable with your own tools ↗
Who's building Varve
Varve is built by Sagi Bashari, former CTO of MyHeritage, where for over a decade he led the company's AWS migration and built the distributed batch-processing systems behind its matching and discovery technologies, scaling them to 39 billion+ records. After years of operating sprawling storage estates at that scale, Varve is the product he wanted. Outcrop shows the engine on public data.
Recent writing
- Buckets and objects are not enough ↗ Why object storage needs a dataset abstraction it doesn't have.
- How intelligent is S3 Intelligent-Tiering? ↗ What S3's automatic tiering can and can't know about your data.
Private preview
Get involved early
Varve's commercial product is in active development and private preview with a small group of design partners. It is not generally available yet. You can apply as a design partner or follow product progress.
Get product updates
Not a fit yet, or not ready? Leave your details and we'll keep you posted as we make progress and let you know when Varve is ready for customers.
Get product updates