About Polign.
The idea behind building Polign is to make agent memory infrastructure lightweight, simple, and private. The software should be capable of running entirely in your own cloud or on your own devices, and the data should never leave your control.
The idea in three steps
The problem
From being an early engineer on AWS S3, to later leading the data orchestration platforms
at Google Cloud Storage, I have always followed one principle: the underlying systems
should remain as transparent as possible, and the technology above it should be where
the intelligence lives.
The rise of agent memory made the contrast stark. Search moved out of the
database and into vector and graph stores, and what an agent remembers became
data you cannot afford to lose. Yet the options on offer forced a bad trade:
run a large cluster of your own just to hold that data, or hand it to a hosted
system that owns its lifecycle.
The insight
An agent's memory and semantic knowledge deserve better than a hosted database with an export path. Over two decades, the cloud storage bucket earned its place as the most durable, most reliable home for data there is. The insight behind Polign: with the right software above it, the bucket itself can be the database, offering the semantics of a traditional database on top of the durability of storage that is already well trusted.
The product
polign_db is that software: a single binary that serves vector, keyword, and hybrid search with typed metadata, and writes everything to a bucket you control in a format you can always read back. The bucket does the heavy lifting of durability; the binary stays small and self-contained, runs comfortably on small machines, and uses local disk only as a cache, bounded by limits you set.
Behind Polign
Anup Talwalkar
Founder and CEO · LinkedIn
As an early engineer on AWS S3, Anup helped define the APIs and how users interact with one of the largest storage platforms in the world, continuing to build products such as S3 lifecycle and event notifications from the ground up.
At Google Cloud Storage, Anup led the data orchestration platforms (Storage Transfer) and built the event-driven transfer framework that powers the near real-time workflows and Cross-bucket Replication services.
At Uber, Anup built Uber/Fx, the dependency injection framework that powers Uber's Go microservices and is now used by thousands of Go developers worldwide.
Reachable at anup@polign.com
Be part of the next chapter.
Tell us what you are building, and we will get back to you.