Problem

Deep learning is the focus of a great deal of machine learning research, but has not become ubiquitous yet in industry because it faces a few huge problems. Today machine learning is intrinsically tied to big data; the more data we have, the better predictions we can make. And as we analyze more data, we can afford to use models with more representational capacity -i.e. ‘big models’, namely neural networks that have millions of parameters and end up being more accurate than any other traditional machine learning method.

Centralizing observations from users presents several challenges. Bandwidth and storage costs can balloon out of control when trying to store images or large amounts of textual/numerical data. Any data center which contains such sensitive information is a prime target for hackers, but has to sacrifice many security guarantees to be capable of accepting incoming data streams from a dynamic set of consumer-owned devices. Moreover once a hacker is inside the walls of a data center, they have access to all the data stored within.

Deep learning means big models, and big models need big data. Given how effective deep learning is, it’s no surprise that more and more companies are centrally aggregating petabytes of data for training these models. But the exponentially increasing number of data breaches is a stark reminder of what can go wrong when one of these data goldmines is breached. Whether it’s corporate espionage, hacking, or misuse of access, billions of records are expected to be compromised in 2019.

This is hardly news, though. Governments around the world are very aware of what happens when data falls into the wrong hands, and so to protect consumer privacy we’ve seen regulations such as GDPR in the EU and the Consumer Privacy Act in California passed this year alone. Much like HIPAA, they make it harder than ever to collect, store, and analyze consumer data.

Solution

We push deep learning models to the end users and train them on the edge. This allows model developers to train on data they’ve never had access to while maintaining privacy, security, and compliance. By ensuring the data never leaves the host, we solve privacy and security concerns. By pushing machine learning to the edge, we can train on much more data than would be feasible to train on in a centralized setting. All of this while generalizing to better machine learning models using federated learning. This type of "decentralized" learning has only been deployed in a couple settings, and the algorithm has always had to be tweaked a huge amount to accommodate the nature of the learning. With our service, we can generalize the powerful algorithm to many machine learning settings.

If you're interested in our work, please check out our codebase or get in touch via email!

Team


Ashwinee Panda

Georgy Marrero

Neelesh Dodda

Nicolas Zoghb

Demo

This is a demo of our iOS app.


This is a demo of the website.

Elements

Text

This is bold and this is strong. This is italic and this is emphasized. This is superscript text and this is subscript text. This is underlined and this is code: for (;;) { ... }. Finally, this is a link.


Heading Level 2

Heading Level 3

Heading Level 4

Heading Level 5
Heading Level 6

Blockquote

Fringilla nisl. Donec accumsan interdum nisi, quis tincidunt felis sagittis eget tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan faucibus. Vestibulum ante ipsum primis in faucibus lorem ipsum dolor sit amet nullam adipiscing eu felis.

Preformatted

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';

Lists

Unordered

  • Dolor pulvinar etiam.
  • Sagittis adipiscing.
  • Felis enim feugiat.

Alternate

  • Dolor pulvinar etiam.
  • Sagittis adipiscing.
  • Felis enim feugiat.

Ordered

  1. Dolor pulvinar etiam.
  2. Etiam vel felis viverra.
  3. Felis enim feugiat.
  4. Dolor pulvinar etiam.
  5. Etiam vel felis lorem.
  6. Felis enim et feugiat.

Icons

Actions

Table

Default

Name Description Price
Item One Ante turpis integer aliquet porttitor. 29.99
Item Two Vis ac commodo adipiscing arcu aliquet. 19.99
Item Three Morbi faucibus arcu accumsan lorem. 29.99
Item Four Vitae integer tempus condimentum. 19.99
Item Five Ante turpis integer aliquet porttitor. 29.99
100.00

Alternate

Name Description Price
Item One Ante turpis integer aliquet porttitor. 29.99
Item Two Vis ac commodo adipiscing arcu aliquet. 19.99
Item Three Morbi faucibus arcu accumsan lorem. 29.99
Item Four Vitae integer tempus condimentum. 19.99
Item Five Ante turpis integer aliquet porttitor. 29.99
100.00

Buttons

  • Disabled
  • Disabled

Form