2023-Management-Sync_Computing

JEFF CHOU: Hi, everyone. My name is Jeff Chou I'm the co-founder and CEO of Sync Computing, and we are helping companies continuously optimize their cloud data workloads. As everyone saw here, in all of these talks, there's a fundamental component to all of these great ML algorithms. It's the data. And moving and storing and computing and transforming the data is a huge cost driver for many, many large companies. And as probably the problem is very clear to most folks, Cloud costs are not shrinking. This is a report by Andreessen Horowitz where you see, in the pink lines, the Cloud costs have actually surpassed on prem infrastructure costs. And this number is not shrinking.

For Sync, what we believe is that the core problem is really at the data engineer. It's really at the folks who are actually deploying and launching these models or these large tables on the Cloud. And this is-- on the left here is what is done today, where a data engineer will be building their code, testing and deploying, and eventually gets launched. And then they might monitor it to see how it's doing, how the Cloud costs are. But there's one missing step. That is the optimization of your code and of the infrastructure. And that's because it's incredibly difficult to do that. It's slow. It's manual, and most people just don't have the time to do it. At Sync, what we're trying to do is complete that bottom loop where we can actually make the optimization of the infrastructure automatic. We want to basically help continuously tune and optimize the infrastructure that you're running on so that you can actually get the cost goals that you're trying to achieve.

The big three takeaways at Sync are, one, obviously, we can help reduce your Cloud costs very significantly. Two, we can also improve the efficiency of your data engineers. Now they don't have to spend time manually tuning memory interconnect instances on Amazon, product pricing, et cetera. They can just focus on the business logic. And three, we can help reduce risk for mission critical workloads. If you have a job that has to complete within 10 minutes-- otherwise, your customers are all going to be upset-- we can help you achieve that as well. Some user results and productions-- we've helped a variety of companies from some of the largest global streaming companies. If you've watched a movie this weekend, you probably used us. From large auto manufacturing companies. And you can see that the savings vary. It depends on the workload itself.

So we go in and do a workload by workload based optimization. And you can see, we can help folks become cheaper, or faster, whatever their business goals are. Under the hood, what's going on is, let's say you run these production jobs. They might run daily, hourly, every minute. What happens is, you run your job on the left here. And once it's done, it outputs a log of what happened in the compute side. We then get that log. We update our model. We then select a prediction based on what your goals are. We then return to you a new configuration such that the next time you run the job, you can achieve the goals that you're trying to achieve.

So what exactly is optimized? What do we actually do? So like I mentioned earlier, we optimize the infrastructure. So we don't actually touch user code. Code is-- it is what it is. We black box it, and we try to bend the Cloud Infrastructure around it to make it the most optimal possible. So for example, there's application tuning. There's Cloud hardware, Cloud economics, scheduling. And so all of these things are changing all the time, not to mention the code itself is changing, or the data size, or data skew itself is all changing. And so you need this continuous optimization to make sure that all of your business goals are achieved.

So about the team. It's myself and my co-founder, Saraj. We're out of MIT and MIT Lincoln Lab. We also have some veterans from the Cloud Infrastructure world, and we have staff and advisors from some of these great companies that you all recognize. In terms of the partnership ask, we've had a lot of success with companies that run large data sets on the Cloud today. So if you're running on Amazon today or using Databricks to fuel all of your infrastructure, you're a great target for us to help out today. And your big problem is, usually, for the most part, it's cost. Your Databricks bill or your [INAUDIBLE] bill is just crazy, and you're trying to reduce it. Especially in today's economy, it's really important and a top line issue for folks.

So right now, we are doing POC efforts to do an initial demo, and then we'd love to roll you all into a more integrated solution. So if this sounds interesting, please feel free to stop by the booth. It'd be great to talk to you all. Thanks so much.