Personal blog of Yzmir Ramirez

In 2016, Serkan Piantino packed up his desk at Facebook with hopes to move on to something new. The former director of Engineering for Facebook AI Research had every intention to keep working on AI, but quickly realized a huge issue.

Unless you’re under the umbrella of one of these big tech companies like Facebook, it can be very difficult and incredibly expensive to get your hands on the hardware necessary to run machine learning experiments.

So he built Spell, which today received $15 million in Series A funding led by Eclipse Ventures and Two Sigma Ventures.

Spell is a collaborative platform that lets anyone run machine learning experiments. The company connects clients with the best, newest hardware hosted by Google, AWS and Microsoft Azure and gives them the software interface they need to run, collaborate and build with AI.

“We spent decades getting to a laptop powerful enough to develop a mobile app or a website, but we’re struggling with things we develop in AI that we haven’t struggled with since the 70s,” said Piantino. “Before PCs existed, the computers filled the whole room at a university or NASA and people used terminals to log into a single main frame. It’s why Unix was invented, and that’s kind of what AI needs right now.”

In a meeting with Piantino this week, TechCrunch got a peek at the product. First, Piantino pulled out his MacBook and opened up Terminal. He began to run his own code against MNIST, which is a database of handwritten digits commonly used to train image detection algorithms.

He started the program and then moved over to the Spell platform. While the original program was just getting started, Spell’s cloud computing platform had completed the test in less than a minute.

The advantage here is obvious. Engineers who want to work on AI, either on their own or for a company, have a huge task in front of them. They essentially have to build their own computer, complete with the high-powered GPUs necessary to run their tests.

With Spell, the newest GPUs from Nvidia and Google are virtually available for anyone to run their tests.

Individual users can get on for free, specify the type of GPU they need to compute their experiment and simply let it run. Corporate users, on the other hand, are able to view the runs taking place on Spell and compare experiments, allowing users to collaborate on their projects from within the platform.

Enterprise clients can set up their own cluster, and keep all of their programs private on the Spell platform, rather than running tests on the public cluster.

Spell also offers enterprise customers a “spell hyper” command that offers built-in support for hyperparameter optimization. Folks can track their models and results and deploy them to Kubernetes/Kubeflow in a single click.

But perhaps most importantly, Spell allows an organization to instantly transform their model into an API that can be used more broadly throughout the organization, or used directly within an app or website.

The implications here are huge. Small companies and startups looking to get into AI now have a much lower barrier to entry, whereas large traditional companies can build out their own proprietary machine learning algorithms for use within the organization without an outrageous upfront investment.

Individual users can get on the platform for free, whereas enterprise clients can get started for $99/month per host you use over the course of a month. Piantino explains that Spell charges based on concurrent usage, so if the customer has 10 concurrent things running, the company considers that the “size” of the Spell cluster and charges based on that.

Piantino sees Spell’s model as the key to defensibility. Whereas many cloud platforms try to lock customers in to their entire suite of products, Spell works with any language framework and lets users plug and play on the platforms of their choice by simply commodifying the hardware. In fact, Spell doesn’t even share with clients which cloud cluster (Microsoft Azure, Google or AWS) they’re on.

So, on the one hand the speed of the tests themselves goes up based on access to new hardware, but, because Spell is an agnostic platform, there is also a huge advantage in how quickly one can get set up and start working.

The company plans to use the funding to further grow the team and the product, and Piantino says he has his eye out for top-tier engineering talent, as well as a designer.

The Pentagon is going to make one cloud vendor exceedingly happy when it chooses the winner of the $10 billion, ten-year enterprise cloud project dubbed the Joint Enterprise Defense Infrastructure (or JEDI for short). The contract is designed to establish the cloud technology strategy for the military over the next 10 years as it begins to take advantage of current trends like Internet of Things, artificial intelligence and big data.

Ten billion dollars spread out over ten years may not entirely alter a market that’s expected to reach $100 billion a year very soon, but it is substantial enough give a lesser vendor much greater visibility, and possibly deeper entree into other government and private sector business. The cloud companies certainly recognize that.

That could explain why they are tripping over themselves to change the contract dynamics, insisting, maybe rightly, that a multi-vendor approach would make more sense.

One look at the Request for Proposal (RFP) itself, which has dozens of documents outlining various criteria from security to training to the specification of the single award itself, shows the sheer complexity of this proposal. At the heart of it is a package of classified and unclassified infrastructure, platform and support services with other components around portability. Each of the main cloud vendors we’ll explore here offers these services. They are not unusual in themselves, but they do each bring a different set of skills and experiences to bear on a project like this.

It’s worth noting that it’s not just interested in technical chops, the DOD is also looking closely at pricing and has explicitly asked for specific discounts that would be applied to each component. The RFP process closes on October 12th and the winner is expected to be chosen next April.

Amazon

What can you say about Amazon? They are by far the dominant cloud infrastructure vendor. They have the advantage of having scored a large government contract in the past when they built the CIA’s private cloud in 2013, earning $600 million for their troubles. It offers GovCloud, which is the product that came out of this project designed to host sensitive data.

Many of the other vendors worry that gives them a leg up on this deal. While five years is a long time, especially in technology terms, if anything, Amazon has tightened control of the market. Heck, most of the other players were just beginning to establish their cloud business in 2013. Amazon, which launched in 2006, has maturity the others lack and they are still innovating, introducing dozens of new features every year. That makes them difficult to compete with, but even the biggest player can be taken down with the right game plan.

Microsoft

If anyone can take Amazon on, it’s Microsoft. While they were somewhat late the cloud they have more than made up for it over the last several years. They are growing fast, yet are still far behind Amazon in terms of pure market share. Still, they have a lot to offer the Pentagon including a combination of Azure, their cloud platform and Office 365, the popular business suite that includes Word, PowerPoint, Excel and Outlook email. What’s more they have a fat contract with the DOD for $900 million, signed in 2016 for Windows and related hardware.

Azure Stack is particularly well suited to a military scenario. It’s a private cloud you can stand up and have a mini private version of the Azure public cloud. It’s fully compatible with Azure’s public cloud in terms of APIs and tools. The company also has Azure Government Cloud, which is certified for use by many of the U.S. government’s branches, including DOD Level 5. Microsoft brings a lot of experience working inside large enterprises and government clients over the years, meaning it knows how to manage a large contract like this.

Google

When we talk about the cloud, we tend to think of the Big Three. The third member of that group is Google. They have been working hard to establish their enterprise cloud business since 2015 when they brought in Diane Greene to reorganize the cloud unit and give them some enterprise cred. They still have a relatively small share of the market, but they are taking the long view, knowing that there is plenty of market left to conquer.

They have taken an approach of open sourcing a lot of the tools they used in-house, then offering cloud versions of those same services, arguing that who knows better how to manage large-scale operations than they do. They have a point, and that could play well in a bid for this contract, but they also stepped away from an artificial intelligence contract with DOD called Project Maven when a group of their employees objected. It’s not clear if that would be held against them or not in the bidding process here.

IBM

IBM has been using its checkbook to build a broad platform of cloud services since 2013 when it bought Softlayer to give it infrastructure services, while adding software and development tools over the years, and emphasizing AI, big data, security, blockchain and other services. All the while, it has been trying to take full advantage of their artificial intelligence engine, Watson.

As one of the primary technology brands of the 20th century, the company has vast experience working with contracts of this scope and with large enterprise clients and governments. It’s not clear if this translates to its more recently developed cloud services, or if it has the cloud maturity of the others, especially Microsoft and Amazon. In that light, it would have its work cut out for it to win a contract like this.

Oracle

Oracle has been complaining since last spring to anyone who will listen, including reportedly the president, that the JEDI RFP is unfairly written to favor Amazon, a charge that DOD firmly denies. They have even filed a formal protest against the process itself.

That could be a smoke screen because the company was late to the cloud, took years to take it seriously as a concept, and barely registers today in terms of market share. What it does bring to the table is broad enterprise experience over decades and one of the most popular enterprise databases in the last 40 years.

It recently began offering a self-repairing database in the cloud that could prove attractive to DOD, but whether its other offerings are enough to help it win this contract remains to be to be seen.

AWS has its Snowball (and Snowmobile truck), Google Cloud has its data transfer appliance and Microsoft has its Azure Data Box. All of these are physical appliances that allow enterprises to ship lots of data to the cloud by uploading it into these machines and then shipping them to the cloud. Microsoft’s Azure Data Box launched into preview about a year ago and today, the company is announcing a number of updates and adding a few new boxes, too.

First of all, the standard 50-pound, 100-terabyte Data Box is now generally available. If you’ve got a lot of data to transfer to the cloud — or maybe collect a lot of offline data — then FedEx will happily pick this one up and Microsoft will upload the data to Azure and charge you for your storage allotment.

If you’ve got a lot more data, though, then Microsoft now also offers the Azure Data Box Heavy. This new box, which is now in preview, can hold up to one petabyte of data. Microsoft did not say how heavy the Data Box Heavy is, though.

Also new is the Azure Data Box Edge, which is now also in preview. In many ways, this is the most interesting of the additions since it goes well beyond transporting data. As the name implies, Data Box Edge is meant for edge deployments where a company collects data. What makes this version stand out is that it’s basically a small data center rack that lets you process data as it comes in. It even includes an FPGA to run AI algorithms at the edge.

Using this box, enterprises can collect the data, transform and analyze it on the box, and then send it to Azure over the network (and not in a truck). Using this, users can cut back on bandwidth cost and don’t have to send all of their data to the cloud for processing.

Also part of the same Data Box family is the Data Box Gateway. This is a virtual appliance, however, that runs on Hyper-V and VMWare and lets users create a data transfer gateway for importing data in Azure. That’s not quite as interesting as a hardware appliance but useful nonetheless.

Microsoft today launched two new Azure regions in China. These new regions, China North 2 in Beijing and China East 2 in Shanghai, are now generally available and will complement the existing two regions Microsoft operates in the country (with the help of its local partner, 21Vianet).

As the first international cloud provider in China when it launched its first region there in 2014, Microsoft has seen rapid growth in the region and there is clearly demand for its services there. Unsurprisingly, many of Microsoft’s customers in China are other multinationals that are already betting on Azure for their cloud strategy. These include the likes of Adobe, Coke, Costco, Daimler, Ford, Nuance, P&G, Toyota and BMW.

In addition to the new China regions, Microsoft also today launched a new availability zone for its region in the Netherlands. While availability zones have long been standard among the big cloud providers, Azure only launched this feature — which divides a region into multiple independent zones — into general availability earlier this year. The regions in the Netherlands, Paris and Iowa now offer this additional safeguard against downtime, with others to follow soon.

In other Azure news, Microsoft also today announced that Azure IoT Edge is now generally available. In addition, Microsoft announced the second generation of its Azure Data Lake Storage service, which is now in preview, and some updates to the Azure Data Factory, which now includes a web-based user interface for building and managing data pipelines.

No matter what cloud you build on, if you want to build something that’s highly available, you’re always going to opt to put your applications and data in at least two physically separated regions. Otherwise, if a region goes down, your app goes down, too. All of the big clouds also offer a concept called ‘availability zones’ in their regions to offer developers the option to host their applications in two separate data centers in the same zone for a bit of extra resilience. All big clouds, that is, except for Azure, which is only launching its availability zones feature into general availability today after first announcing a beta last September.

Ahead of today’s launch, Julia White, Microsoft’s corporate VP for Azure, told me that the company’s design philosophy behind its data center network was always about servicing commercial customers with the widest possible range of regions to allow them to be close to their customers and to comply with local data sovereignty and privacy laws. That’s one of the reasons why Azure today offers more regions than any of its competitors, with 38 generally available regions and 12 announced ones.

“Microsoft started its infrastructure approach focused on enterprise organizations and built lots of regions because of that,” White said. “We didn’t pick this regional approach because it’s easy or because it’s simple, but because we believe this is what our customers really want.”

Every availability zone has its own network connection and power backup, so if one zone in a region goes down, the others should remain unaffected. A regional disaster could shut down all of the zones in a single region, though, so most business will surely want to keep their data in at least one additional region.

 Microsoft announced a new product in its Azure line-up in preview today that will make it easier for developers to build event-based applications. The Azure Event Grid makes events (like uploading a picture or video, clicking a button, updating a database, etc.) first-class Azure objects. Event Grid complements Azure Functions and Azure Logic Apps, Microsoft’s existing serverless offerings. Read More

 Since the dawn of Azure, Microsoft has talked about how enterprises can benefit from a hybrid cloud approach — that is, using the public cloud while still running some of their applications in their own data centers. Even today, Microsoft says that 80 percent of the companies it talks to still want to use a hybrid cloud approach and to help them move to its cloud services, the company… Read More

 Companies like Microsoft, Amazon and Google continue to compete fiercely in the area of cloud services for consumers, developers and enterprises, and today Microsoft made its latest moves to lay out its bid to lead the race, while also launching a new mission to position itself as the cloud provider you can trust. Microsoft announced it would build its first Azure data center in France… Read More

 Microsoft’s CTO for Azure (and occasional novelist) Mark Russinovich is extremely bullish about microservices. In his view, the vast majority of apps — including enterprise apps — will soon be built using microservices. Microsoft, with its variety of cloud services and developer tools, obviously wants a piece of that market. With Service Fabric, the company offers a service… Read More

 MongoDB, the company behind the eponymous open source database, is launching Atlas today, its third major revenue-generating service. Atlas is MongoDB’s database-as-a-service offering that provides users with a managed database service. The service will offer pay-as-you-go pricing and will initially allow users to deploy on Amazon Web Services (AWS), with support for Microsoft Azure… Read More