 Whatever you work, this is a fact of life. Whatever data you receive is probably wrong. Okay, so the first thing you need to do is check it. You need to make sure that the data is clean, because if the data is bad, then your result will be bad. So the first thing we do is we have a series of checks, both rule-based and algorithmic-based, to make sure that the data that we feed into our solution is clean. Next step is we build a rough solution. Okay? It doesn't need to be perfect. It doesn't need to solve all the issues. It doesn't need to be highly profitable. It just needs to be something that we can use as a starting point, a seed. And we use some heuristic to generate this. And then we improve. So we improve on two parts. We first improve on operability. All these operational constraints that I listed here, we try to fix these first, because it doesn't matter. You could have the most profitable schedule on paper. You could fly it the best times, but if your schedule is not operable, that means it won't fly. Come day of operation, you'll use too many gates. You'll violate curfews, things like that. So you'll have issues left and right. So you will end up losing money. So the first thing we need to do is focus on feasibility. Make sure that the schedule can actually fly. Then we focus on profitability. Then we retime, we refleet, do whatever we can to make sure that you as a passenger actually want to fly this airline. And we use too many algorithms to do that. We use MIP, where we extrapolate the... MIP stands for mixed integer programming. So we basically extrapolate the problem into a mixed integer programming model, run it through existing solvers, and then extrapolate that back out into an actual business problem. Another approach is network optimization, also referred to as neighborhood search by our team. Neighborhood search, basically, you start with a solution. You make incremental changes in the neighborhood. You evaluate which one's the best. You choose that one as your next starting point, and then you take it from there. The underlying construct behind all of this is a network. So you saw that in the company video. We're all about networks. So the underlying network that we use is a time-space network. The nodes will be planes in certain cities at certain times. And then we can use arcs to denote flights. We can use arcs to denote ground time at a gate. You can use arcs to denote connections for the passenger connections. And the goal is to optimize the flow through this network. Again, ensuring that you're honoring all of your operational constraints. Our solution, branded, is SkyMax. So I actually have some SkyMax team members here. They're sitting right there. Hey, guys. So they worked on it. They're all very proud of it. So one part is actually solving the business problem. You need to have this complicated engine that can solve the problem. But who's going to be using this? There's going to be basic users who aren't OR experts or masters in industrial engineering. There are just basic people who want an optimizer to solve their problem. They're like, I fly to these cities. I fly to these markets. They want an easy application. So it's not just solving the business problem. It's wrapping it in a solution that the user can actually benefit from. So that's what SkyMax does. It solves for flight times. It solves for fleet types. It creates optimal routings. It can build a schedule from scratch. So no other solution in the marketplace can do that. And it even has the ability to detect issues with your data and fix those issues. If you don't mind my asking, what I'm curious about is you guys have been able to come along and do this. The major airlines have been working on this for a long time. Yes. So can you talk a little bit about what you were able to do. It looks like recently in the last 10 years you've really moved this along as opposed to say what America and the United have been doing. Right. So if you're familiar with our competition, Saber is actually your number one competitor. And the ex-president of Saber actually went ahead and said that clean sheet scheduling, that's what it's referred to, is creating a schedule from scratch, is not feasible. So I would never shoot myself in the foot like that. But we went ahead and did it. So I can't really disclose what exactly we did, how we solved it. But they use techniques like column generation and not any mix of related algorithms to kind of solve the problem. And true clean sheet scheduling, so I don't want to boast ourselves too much. Clean sheet scheduling involves not only determining like creating optimal flight times and optimal schedules, but it's also determining the optimal number of flights, which is I want to fly these 10 cities, or maybe I want to fly. Where do I fly? How many flights do I have? The first step, that is currently an input into Skymax, but true clean sheet means I'm an airline, I only have 10 planes. Where do I go? When do I go? So we actually don't do true clean sheet in that sense, but it's as part of our suite, that is the next step. And the only way we can do that is actually to decompose the problem. You need to make simplifying assumptions in order to, say, solve the frequency plan to determine the optimal number of flights between any two cities. And then once you've settled on that, then you use that as your assumption for the next step. So you actually can't solve everything at once, at least based off the computing power that we have now. So the only way you can do it is if you make simplifying assumptions. But if the result is still positive, which you'll see in the next slide it is, it's worth it. Hopefully that answered your question. I guess this is all deterministic. There are all kinds of uncertainties like... It's actually nondeterministic. So you're not guaranteed to get the same solution every time. So you have all passengers, arrival time, departure time, are you accounting for those? True. Sorry, did you repeat that? Are you accounting for the uncertainties, for example, number of passengers or arrival, departure time, weather conditions? As part of our suite, we have products that focus on that, but during the construction of the schedule, we don't focus on that. So we make... Is climax deterministic optimization? No, it's non-deterministic optimization. So what is the uncertainty? What aspects are... The uncertainty depends on the way the model is built. What variables are... What are the uncertain variables? So to clarify, I'm actually not from the OR team. There are details of the optimization agent, which I'm not that familiar with. I'd be more than happy to connect you with the OR experts behind this. You're way more qualified than I am in this area, but I don't want to lead you the wrong way or tell you the wrong answer. Any other questions not related to the agent? So you might be wondering, yeah, this application might be great, the agent might work, but what is the real value to the end user? So we did this case study with Southwest Airlines. So to give you a rough estimate, they have about 700 planes, almost 4,000 flights a day. And so they're the ultra mega airlines. So they're one of the largest airlines in the world. And the benefits that they saw, at least on paper, came to about a 2-3% increase in revenue. And that translates to roughly $580 million annually. So it translates to about a 10% increase in profit, which is $2-3 million a day, or $1-2 million a day. So for them, the ROI was tremendous enough to actually either buy the software or continue pursuing this project. So Southwest was actually a developmental partner that funded the development for this and helped us build this from ground zero. What is the cost of the software? It varies. Interested? Yeah. So it varies. And you call us in bikini a special class. Yeah. So it varies based off the size of the airline. So as I mentioned, Southwest is a large airline, so the value could be on scale of hundreds of millions of dollars annually. So they are willing to spend millions of dollars a year on it. And for a large airline like that, millions of dollars is really not that much. But for smaller airlines, maybe they only have 50 planes or 60 planes, the value will be much less because the impact is much less. So we might charge at least a third or fourth of the price for the same solution. And that's pretty common in this space. So right now, I only talked about Skymax, but our vision and our goal is to solve other problems in this domain. So we know that you can have a perfect schedule and fly it, but on the day of operations, things will go wrong. There are weather issues that are unaccounted for when you optimize. So what we're trying to do is we simulate what happened historically and gauge the on time performance of that schedule. And then ultimately, make changes to the schedule to basically make the schedule more robust to absorb those delays that actually happen in real life. Skyplant would be the precursor to Skymax. It would optimize the frequency plan. Frequency plan is the number of flights between any two cities. It would optimize that frequency plan and then feed that into Skymax, which would then build a schedule on top of that. So those two combined would actually be true clinchy schedule. Skycast is the main purpose is to actually model passenger choice. What exactly does a passenger do? Whenever they're searching for tickets, what things are they looking at? How do they choose between a similar flight for a different airline? How do they choose between business or economy? So to do real passenger choice modeling. And Skyworks is our schedule editor. Once you have a schedule, you do need a tool and this probably will be the most popular of all of our products. You need a tool to actually manage and edit the schedules. So while larger airlines can actually build schedules manually so they might not need Skymax, smaller ones will need an editor because you always need to make changes and then publish those changes out into GDSs so that people can actually buy those tickets. So Skymax, SkySim, we've actually already built but are constantly improving. SkyPlan, Skycast, and Skyworks, we're just starting right now. And we're looking for e-reminds to help us solve these problems. Thank you. What is the routine or the workflow of using this order? I'm an airline company and I have a suite. How frequently do I do that calculation? Once per day, once per week, or what? Excellent question. So generally what airlines do, so it depends on the airline. So Southwest, for example, they take the calendar year. So every airline that operates daily needs to have a schedule for that day. So how do they build that schedule for every day of the year? So what Southwest does is they take the calendar year