Data integration without a schema meeting first
Most integration tools ask you to declare the shape of the data before you are allowed to look at it. DataLens inverts that: land the rows, let the platform profile them, and decide what the shape should be once you can see what actually arrived.
What lands, and what happens to it
Every source in DataLens produces the same thing: a dataset in the catalogue, with a profile attached. That profile is computed on arrival — column types, null rates, distinct counts, distributions, candidate keys and the quality problems that fall out of them.
Because the profile is generated rather than declared, the first useful conversation about a new source happens minutes after connecting it, not after a modelling workshop. Ari has the profile in front of it, so you can ask what is wrong with the data before you have written a single transform.
Nine ways in
Each one lands in the same catalogue and behaves the same from that point on.
CSV & Excel upload
Drag a file in and start working. Profiled on arrival, no schema declared up front.
Database
Read from your existing relational database and pull tables in as datasets.
REST API
Point at an endpoint, map the response shape, and land it as rows.
FTP & SFTP
Collect the drops that still arrive as files on a server.
CDC sync
Change data capture, so the dataset follows the source instead of ageing.
CRM
Bring customer records across without exporting them to a spreadsheet first.
Data lake
Read from — and write back to — object storage you already own.
Documents
PDFs and documents stored whole, for retrieval and generation rather than rows.
AI model providers
Your own model credentials, held as a connection like any other source.
What you get on arrival
Automatic profiling
Types inferred from the values rather than the header, with null rate, cardinality and distribution per column.
A working copy
Transforms run against a working copy, so the bytes that arrived stay intact and recoverable no matter what you do next.
Catalogue metadata
Row counts, column descriptions and a data dictionary that is generated from the data and then edited, rather than authored from nothing.
Immediate quality signal
Duplicates, type mismatches, outliers and suspicious nulls are surfaced as findings, not left for a dashboard to notice later.
PII detection
Columns that look like personal data are flagged at ingest, so a governance decision can be made before the data spreads.
Lineage from the first step
Where the dataset came from is recorded when it lands, not reconstructed afterwards from logs.
Data integration questions
Do I need to define a schema before connecting a source?
No. DataLens infers types and structure from the values that arrive and shows you the profile. You define a target schema later, in the modelling stage, once you know what you are actually working with.
Can I keep data in my own storage?
Yes. The data lake connector reads from and writes back to object storage you own, so the rows can stay in your account.
How do incremental loads work?
The CDC connector captures changes at the source so the dataset tracks it over time, rather than being replaced by a full reload on every run.
What happens to a file that fails to parse?
It surfaces as a finding against the ingest rather than a stack trace. Malformed rows are reported with their position so you can decide whether to fix, quarantine or drop them.
See it on your own data
DataLens is in private beta. Bring a file, a database or an API and work through the whole lifecycle in one sitting.
Request beta access