Jeff Chou and Suraj Bramhavar are co-founders of Sync Computing, a deep tech startup spun out of MIT that has figured out how to automatically reconfigure and reschedule big data and machine learning jobs, making the cloud easier, faster, and cheaper.
Welcome to the cloud, where it’s easier to derive insights from big data, innovation is accelerated, and greater opportunities for collaboration live alongside the ability to scale up or scale down IT resources as necessary. And say goodbye to the costs associated with racks of in-house server hardware and software updates, among other things. Everyone is either on the cloud or wants to go there. According to a report by Andreessen Horowitz, over $US 100 billion is spent each year on cloud services. Jeff Chou and Suraj Bramhavar, the co-founders of MIT spinout Sync Computing, believe that half of that money is wasted, and they want to do something about it.
Outsourcing infrastructure to the cloud might seem like a recipe for simplicity and cost-effectiveness. But the reality of processing data and machine learning workloads on the cloud is getting more complicated and expensive by the day. Amazon Web Services alone offers an evolving menu of 400 different machines with hundreds of settings and fluctuating prices. Add in the various parameters associated with large-scale data processing frameworks like Spark, and the number of possible deployment options becomes impossibly large. That's hundreds of machines, thousands of tasks, and trillions of combinations.
Determining cloud infrastructure settings to optimize job cost and speed is neither practical nor physically possible for cloud developers at such large scales—until now
“With today’s constantly expanding use of large-scale cloud computing, it’s not uncommon for a company to have a dozen engineers responsible for managing up to 10,000 data pipelines per day. Determining cloud infrastructure settings to optimize job cost and speed is neither practical nor physically possible for cloud developers at such large scales—until now,” says Chou.
Partnering with Sync, businesses can automatically choose cloud infrastructure for their big data and machine learning jobs based solely on the things they care about: cost and time. Sync also makes the cloud faster and cheaper without touching a company’s source code. “We’ve basically taken a bunch of PhDs brains and knowledge and compressed it into an easy button,” says Bramhavar. Whether “optimal” to you means the fastest, cheapest, or somewhere in between Sync provides the appropriate settings for your needs.
But what about serverless? Yes, the big cloud providers are rolling out options to allocate machine resources on demand, taking care of the servers on behalf of their customers. It does make things easier for programmers and businesses, but cloud providers aren't particularly interested in making jobs run faster or cheaper. If you’re using serverless, you’re probably paying more for the convenience. Sync, on the other hand, provides the simplicity of serverless while still optimizing for cost and performance and allowing businesses to seamlessly switch between the two metrics.
Chou and Bramhavar met while working on a special committee convened to study advanced computing at MIT's Lincoln Laboratory. Out of that effort, they published a paper in Nature's Scientific Reports demonstrating that it was possible to harness nature's desire to minimize energy using a standard electronic circuit, proving that a physical system could solve optimization problems 1,000 times faster than GPUs using massively manufacturable electronic circuits. Think of it as programming with the laws of physics.
“We come from a hardware background, so we understand how expensive it is to build hardware and how immutable it is. We’ve built one piece of hardware that works for one problem that everyone has,” says Chou. That problem is scheduling, otherwise known as resource allocation at a digital level. The result is a radically new way to accelerate jobs, lower costs, and decrease our computing carbon footprint on the cloud. And the technology is agnostic to software platforms, cloud providers, and hardware types, making it perfect for today’s constantly changing cloud environment.
It's a game-changer for any company that processes machine learning or data workloads on the cloud or wants to go to the cloud and cares about performance and cost. One of Sync's first customers is Duolingo. With more than 40 million active monthly users, the world's most popular language-learning platform chugs through terabytes of data on the cloud, which makes optimizing processing performance critical. With the help of Sync Computing, Duolingo was able to reduce the cost of daily data jobs on the cloud by 50 percent.
According to McKinsey the average company staffs about 35 percent of its cloud needs in-house. Most want to bring that number up to 50 percent by 2024, which means organizations will be looking to hire or reskill at least one million new cloud developers over the next few years. But hiring or training PhD-level talent that can do data on the cloud is expensive, and cloud-service giants have a habit of poaching top cloud architects.
We were just named to STEX25 last month, and we’ve already had several face-to-face meetings with big corporations, one of which has resulted in a new pilot program
Sync Computing addresses the intertwined talent, cost, performance bottleneck in a novel way. And with the help of MIT Startup Exchange and MIT Industrial Liaison Program, Chou and Bramhavar are expanding their client base. “We were just named to STEX25 last month, and we’ve already had several face-to-face meetings with big corporations, one of which has resulted in a new pilot program. ILP and the Startup Exchange work together to pre-filter vetted introductions, engendering trust on both sides, improving the signal-to-noise ratio, and helping with business development,” says Chou.
Corporations get bombarded with so-called cloud innovations on a regular basis, but Sync is different. Chou and Bramhavar bring unparalleled knowledge to the space. Scheduling, resource allocation—it’s a hard problem. Everybody needs it, but until Sync, nobody has been able to bridge the gap when it comes to the cloud. “No one else is doing the level of analysis, optimization, and automation that we're doing,” says Chou. “What we’re building utilizes a fundamental breakthrough.” Sync does the hard math, everyone benefits.