Introducing the Nearist VSX Cloud



With Nearist and just 500 milliseconds, you can:

Compare 1.7 billion documents


With Nearist and just 500 milliseconds, you can:

Search 125 million images


With Nearist and just 500 milliseconds, you can:

Accelerate drug discovery by comparing 2 billion molecules



Nearist greatly accelerates big data searches through a revolutionary new hardware platform specifically engineered to handle the computationally demanding task of performing Nearest Neighbor Search on vector representations of content—enabling your search routines to deliver results several orders of magnitude faster than anything else on the market.


  • Step 1
    You have content that is searched frequently–documents, images, products, etc.

  • Step 2
    Convert your content into feature vectors

  • Step 3
    Store the vectors in a Nearist appliance

  • Step 4
    Nearist accelerates your similarity searches

  • Step 5
    Results are returned in less than a second

  • Truly revolutionary performance.

    Forget best-in-class. Nearist has invented a class of its own by taking your search queries to a whole new level of speed. Imagine processing a 420GB data set within about 700ms. That translates to 1.7 billion documents, 125 million images, or 2 billion molecules—all in the blink of an eye. Literally. This isn’t just faster, this is game-changing performance that will have a significant and immediate impact on your operations.

  • The capacity to meet any need.

    With flexible data capacities up to 420GB per server— plus thousands of calculation engines that can process up to 16 queries in parallel with sustained throughput of up to 1,000 queries per second— Nearist practically eliminates scaling challenges since a single server is all most companies will ever need. Yet if your company requires larger data set processing, Nearist makes it easy to connect multiple systems for virtually unlimited data capacity. And with built-in API libraries that support a wide range of applications, Nearist offers easy plug-and-play installation that easily integrates into your existing system.

Significant cost savings.

With its game-changing speed, dense parallelism and vast data capacity, a single Nearist-powered server does the work of up to 100 GPU cores—or even 1,000 commodity CPUs. That adds up to significant cost savings, not only in hardware and maintenance, but also in floor space and power usage.

1 Nearist server 4,000 GPU cores 10,000 commodity CPUs64
Server floor space required 1 Nearist server8 sq ft 4,000 GPU cores240 sq ft 10,000 commodity CPUs642,400 sq ft
Annual power consumption 1 Nearist server$2,400 4,000 GPU cores$74,000 10,000 commodity CPUs64$740,000
Combined savings vs. GPU/CPU 4,000 GPU cores~30x less floor space & power consumption 10,000 commodity CPUs64~300x less floor space & power consumption

* Appendix



By going beyond simple keyword and metadata searches, Nearist enables concept searching—an advanced search technique that uses a feature extraction algorithm to learn the relationships between words based on context. This not only delivers more thorough and accurate searches, it allows businesses to pinpoint very specific information among enormous amounts of data in a tiny fraction of the time.

Here are just a few examples:
Law firms can practically eliminate the tedious task of searching through thousands of pages of discovery or emails to quickly find all information relevant to a case.

Customer service departments can more easily diagnose and resolve problems by searching their archives for similar issues to reference.

Researchers can quickly find related documents that may not turn up in traditional keyword or metadata searches, due to differences in word choice or spelling.




Bioinformatics has revolutionized the pharmaceutical industry by giving it the tools to shift from a trial-and-error approach to drug discovery, to one based on our newfound understanding of protein structure, or even searching for structurally similar molecules to existing medications. While these approaches are very promising and have already led to new therapies, they also present some unique challenges—most notably the vast amounts of data that must be analyzed before a suitable drug candidate can be developed and tested.

But with the ability to search and compare up to 2 billion molecules in just half a second, Nearist promises to significantly speed up the revolution in drug discovery that’s already taking place.



From identifying visually similar content among a sea of images, to detecting copyright infringement in music or photography, Nearist enables an incredibly wide range of media matching possibilities. And since feature vector searches look for similarly parsed data instead of often incomplete or inaccurate metatags and keywords, you get significantly more complete and accurate results in a fraction of the time.

Just some of the possibilities:

  • Image-matching searches on the web or on stock photo websites
  • Automatically scan websites to ensure stock photography clearance
  • Music copyright infringement based on similarities in melody or passages
  • Song identification without embedded metatag info, such as a live concert recording


Nearist’s Vector Search Accelerator cards are the muscle behind our fast similarity search. Unlike a GPU, no special hardware programming or optimization is required to leverage its performance. VSX cards are housed in servers in our datacenter, and are accessed using our simple web API for loading and querying your dataset.

Massive Parallelism

  • Each VSX card performs up to 800 billion feature comparisons per second
  • A single appliance holds up to 10 VSX cards, all tearing through your dataset in parallel at up to 8 trillion feature comparisons per second!

Plentiful Storage

  • Each VSX card contains 42GB of onboard memory for your  dataset – 75% more than a Tesla K80.
  • A 10-card server provides 420GB of memory, enough to  hold several billion feature vectors (depending on their size).


Nearist’s hardware is made available as a cloud offering. A reservation and our simple web API are all you need to immediately begin using a Nearist-powered appliance. Check out our code examples on GitHub, and contact us to start your reservation. Nearist’s VSX cards are hosted in our datacenter, and are made available for either on-demand hourly use for experimentation, or by contract for a long-term reservation.

Explore our GitHub profile to see how easy Nearist is to use.

With next-level hardware that promises to drastically accelerate your feature vector searches, Nearist is ready to literally transform big data—and open up entirely new opportunities that simply weren’t achievable until now. So imagine the possibilities, then contact us for more information and to receive our white paper.

Contact Us

Reach out to us and we’ll be happy to send our whitepaper and talk more.