Know what is where
An index and a dedupe pass, so the system does not read every byte on every run. The transfer is the last step, not the first.
Indexedknow what exists before moving it
Build
AI
MoreSpace
MoreSpace moves photos, video and documents between a phone and the cloud accounts a person already has. Storage is bought per account. One is full, another is nearly empty, and both are being paid for.
Moving that by hand is slow enough that people never do it. At the volume this runs - the infrastructure looks after on the order of 5 petabytes - the problem is knowing what exists where, before a transfer starts.
- Know what is where
- Move only what should move
- Treat the files as private
Move only what should move
New content is watched in the background. What already lives elsewhere stays put. The rest goes to the account that has room.
Dedupemove only what is not already there
Treat the files as private
Every movement is authenticated, authorised and encrypted in transit. That is the condition for carrying the files people actually keep.
5 PBprivate files, checked on every move
The build
Indexing before copying
At this scale, reading everything to decide whether to move it does not finish. The analyser shows where the space went - duplicates, files nothing points at - so a person can see it, and the scheduler can act on the index rather than on a guess.
The pipeline has to be quiet
Background sync that surprises someone is a support incident. Filtering what is already stored, then moving the remainder, is how the product stays something people leave running.
A model inside the pipeline
The model sits in the decisions the pipeline already has to make: what looks like a duplicate, what can wait, what is worth showing. It is not a chat box beside the product. It is how those calls get made, which is the only place a model helps this job.
Private files, ordinary rules
There is no later pass where security gets added. Every movement is authenticated, authorised and encrypted, because the product exists to carry private files between accounts. A faster sync that was careless about that would not have been deployable.
How we worked
What moving data asks of us
Scale as an indexing problem
The interesting number is not how fast a file moves. It is knowing the estate well enough to move less of it.
Security on the default path
There is no version of this where the files move first and the policy arrives in a later release.
Intelligence at the decision
A model that describes the product does not move a file. A model in the dedupe and the schedule does.
Technology
What it is built on
- C#
- MYSQL
- Flutter
- Python
- Go Lang
- Elastic Search
More work
More AI and Build work
SecBoard
The command centre for intelligent vulnerability management.
- C#
- MYSQL
- Go Lang
Need something like this built?
Tell us what you are trying to ship and an engineer will reply within one business day.