We run the index so you do not
One retrieval layer, one number to hold us to, and nobody on your team paged at 3am.
Nodeform started in 2023, after the two founders spent a year running a vector index nobody had budgeted for. It worked at ten million embeddings and fell over at two hundred. The fix was never clever — it was sharding, recall harnesses and a rebuild schedule, week after week. We decided that work should belong to one team, not every team. So we run it, we publish the latency, and we put a number in the contract.
Our Values
Reliability First
A retrieval layer should be boring. Boring is the whole product.
We don't ship
what we wouldn't use.
Our Journey
Today we hold 12 billion vectors across nine regions and answer roughly 40,000 queries a second at peak. The index has never been the interesting part. That is the point.
Meet the team
The people who get paged so your on-call does not.
Meet the team
The people who get paged so your on-call does not.
Ready to stop running
your own vector index?
Read the quickstart, or talk to the engineer who will run your index.
