Polygrid A Search & Discovery EnginePowered by Deep Learning and PyramidDB

Experience it yourself

Try out these Polygrids:


More than 170,000 pieces of art from wikiart.org.

Organized by visual similarity using Deep Learning.

View Collection »


More than 200,000 pieces of art from Dutch Rijksmuseum Open Data Collection.

Organized by visual similarity using Deep Learning.

View Collection »


More than 13,000,000 images from the ImageNet Dataset.

Organized categorically using Treemapping, then by visual similarity using Deep Learning.

Note: We've omitted some controversial and poorly imageable synsets, but viewer discretion is still advised. Read more about the ImageNet cleanup effort.

View Collection »


More than 400,000 items in Amazon's Home & Kitchen category as of July 2014. From Julian McAuley's Amazon Product Data.

Organized categorically using Treemapping, then by visual similarity using Deep Learning.

View Collection »

Fashion Products

More than 40,000 fashion products from the Kaggle Fashion Product Images Dataset.

Organized categorically using Treemapping, then by visual similarity using Deep Learning.

View Collection »

Explore with ease

Polygrid gives you an unprecedented ability to search & discover catalogs, from the small to the truly gargantuan.


Find not just the needle, but also what you didn't know you were looking for.

Polygrid can show search results in the context of similar-but-missed items, as well as all by themselves.

... and discover

Glide through entire regions of items and drill down instantly wherever your curiosity takes you.

As you zoom in, you'll discover more and more items that are unique yet broadly similar in nature.

How it works

We use Machine Learning to organize the items into a neatly arranged grid, where things that are similar are likely to be next to each other.

Then we let the most important items percolate up a stack of smaller and smaller such grids, creating a pyramid of exponentially smaller layers, until we have a suitable starting point for your navigation at the top.

We store this potentially enormous catalog in a purpose-built PyramidDB for incredibly fast serving.

Whereas traditional SQL databases store data in rows, and typical NoSql databases store data in variations on columns or graphs, the Pyramid Database stores the data in a pyramid-shaped stack of grids.

This enables instant serving of any part of the entire catalog, and affords a unique search experience where instead of a small handful, we can put thousands-to-millions of search results directly at your fingertips.

This technology is what powers Polygrid's unique ability to let you both scan around quickly and drill down anywhere in the entire catalog as well as in search results.


To break free from the list-and-paginate prison

Catalogs are typically browsed the way their database allows for: you can search, you can list, and you can paginate.

But you can't really move around in them.

So we set out to build a catalog exploration experience that is as natural as moving around in the physical world, while taking full advantage of the virtual one.

Something intuitive. Immersive. Instant.

No more click-and-wait. No more pagination exhaustion. The matrix we were promised, not the list we got stuck in.

The result of this effort is Polygrid, a revolutionary navigational paradigm. Here is what really sets it apart:

  1. Intuitive exploration

    Scrolling in a list or clicking next just gives you 'more'.

    Drilling down in a Polygrid gives you more of that.

    Every step you take refines what's revealed. Just follow your nose. For once the computer works with your human intuition, not against it.

  2. Freedom of movement

    With a traditional interface you're confined to going in essentially one direction.

    In a Polygrid you can go left, right, up, down, in, and out. Then do it all again just for the joy of exploration.

    And you can move quickly. The rate at which you can explore a Polygrid far exceeds the pace at which you can realistically cover catalogs using ordinary means.

  3. Completeness

    With search-and-list systems it's hard to know if you've actually seen all they've got.

    In a Polygrid the entire catalog is always right there in front of you.

    And your search queries don't have to be perfect. Just a keyword or two to point you in the right direction. Then drill down and have a look around.

Have a catalog?

Do you have a catalog you'd like to see in a Polygrid?

Send us an email: hello@polygrid.com

Want to work on something exciting?

We are always looking for smart people who get things done.

We're especially interested in candidates with notable skill & expertise in deep learning, search, devops, and enterprise sales.

Send us an email introducing yourself: careers@polygrid.com

About us: we're small, profitable, 100% remote, strongly typed where possible (Python for DL training, Scala, Java and TypeScript for everything else), and offer competitive comp packages.