 in the world of operation research, Ravin Raoja, and so he has some unique experience how to bring such kind of achievements from science, from academics, to industry and to implement them and to help really solve real solutions in the world. So I think therefore we are here and just to share how we are doing it and to explain how to do it in the most efficient way. This is just our journey how we grow, so as I said we just started as some kind of from scientific part of from science, so it's in 2000 innovative scheduling has been established, then we grow, we grow from industrial part, we implemented some solution in railroad, tracking, mining, industry, airline industry and also we established different offices in US, Armenia, in Australia and as you see it's growing very fast and now also we are expanding our presence in Europe, in CIS countries and also in Asia-Pacific part. It's very important that you understand that as long as we have we need to use some scientific achievements, operation research, cutting-edge technology so it's very important to have people who can implement it and if you see so we have 50 plus PhDs in our company who helps us to build such kind of successful solutions for different industries and also we are dealing with multinational company with multinational cultures, mix of cultures but we are capable to really bring these multinational cultures in managing in most efficient way and therefore I think it's key of our success. As I said we need to have hybrid or mix of different technologies, fields like transportation and logistics like operation research, computer science, project management, data analytics, information technology and you need to, it's some kind of interdisciplinary field so you need to mix it and really in order to build more or less successful solutions for the real industry in transportation. So I think my colleagues will describe real cases from airlines and railroad but also we have presence in traffic industry and also mining industry and if you see how many types of expert has experienced such kind of solutions it's really basic. So I would like to give floor to Shama so he will describe or represent the airline case study. I think it will be very interesting. Thank you. Okay so this airline case study is actually not only the first one that we did as a company but it's actually one of the most important and difficult problems that an airline must solve. So the business problem. I don't like reading slides but I feel like I must read this one because it's very important. The business problem is to maximize an airline's total network profitability by determining the best flight departure time, the best fleet type and the optimal routings to minimize the number of aircrafts. That's a mouthful so let me walk you through that. Let's look at this example right here. So we have an A330, it leaves GFK at 1 a.m. it's an overnight flight and it gets into Moscow at 5 20 p.m. the next day and then it sits on the ground for about an hour and then it comes to Armenia. It came from New York all the way to Armenia. Can anybody guess what airline operates this route? This is based off the real airline. Half-right or was it from the airline? Air Force. Air Force? Correct. I have actually flown a route very similar to this four times coming to Armenia and I can assure you that it has nothing to do with how awesome I think Air Force is. It's absolutely to do with the fact that it was leaving New York where I lived and it was coming to Armenia where I was going and it left at a time that I wanted to leave and it arrived at a time that I wanted to arrive. So you can see how important this is right now why did they choose an A330? Why that fleet type? Why not an A380? Why not pick a bigger plane? Why go with an Airbus at all? Why not go with a Boeing 777? I mean the airline larger airlines have large fleets so they can assign any plane to this flight. Or maybe it wanted to be more frequent the flights. How would the A380 be less frequent? So in capacity right? So larger planes can have more people. Great. Why one? Why leave at one? Why not leave a little bit earlier and maybe I get into Moscow where I'm new and catch my lunch meeting? Why not leave later? Why get into Moscow around 5 p.m.? I don't have anything to do that day. I would rather sleep in in JFK, spend more time partying in New York and then catch and sleep through my flight and getting to Moscow and then do whatever I want. Or what about this plane? Why does it have to go to Armenia next? Why not London? Why not Kiev? We have all these different options, all these different routes, all these different decisions that an airline has to make. Why did they choose this one? As you can see and this I picked a problem with three fleet types, two flights and two departure times. But a large airline with hundreds of planes, thousands of markets, you could see the problem gets to be very large. So optimal flight times, that's actually one of the biggest challenges that an airline must solve. So everybody has their own time of day preference, right? Or maybe even time of week. Sometimes I want to fly Friday so I can meet up with my friends over the weekend and then come back on Sunday, right? And I want to fly Friday after work so I don't take a day off. And I want to fly Sunday late at night so I have that Sunday. So depends on where you're going, JFK to Moscow. So if I'm leaving New York I have my own time of day preference. Maybe I want to leave early, maybe I want to leave later. So the current that you see here is the preference by time of day. So the demand varies by time of day. The chart on the right is a similar curve but it's the curve for Moscow to Yerevan. Now as you can see this is the same plane that operates both flights sequentially. So here you can say, oh it's easy. Well the peak is right here so let's put the plane right there, okay? Well if the plane... Can you see that? Yeah. So while the peak is right there so let's put the plane somewhere around here. Look at that. I got all these people going from JFK to Moscow. But this plane, it's going to get to Moscow and then it's going to go to Armenia. So well if it leaves JFK at this time and it'll get to Moscow this time and then for the people going to Moscow to Armenia it's I didn't hit this peak at all. So maybe I shift this later. Okay I got that peak. Wait what happened here? So I lost the demand on the JFK to Moscow but I got it in my Moscow to Armenia. So you can see just simply this two-market problem, one departure time, can have a major impact on what you really get as your demand. And that's just the departure time. There's also the fleet type. So the gentleman in the back excellently mentioned that you want to match capacity to demand. So the last thing an airline wants are MD seats, right? MD seats, they just take up space. I know we like them because we can strike their legs, lay down, but an airline hates them because that's money that they're not making. So what they want to do is if they have 70 people flying on that flight, they want to put a plane with exactly 70 seats. So they want to minimize bill and minimize costs. Connections. This is a critical factor and this is what makes this problem extremely complicated. So what do you buy flights? Do you buy flights? Yes? Yes. Do you buy, when you go online, do you say you search for a flight or do you search for a trip to go from point A to point B, which might touch most flights? Itinerary. Exactly. So what we really buy, the money that really comes from is not from the flights, it's actually for itineraries. I could, there could be one flight operated by one plane, I could get on another flight operated by another plane. So the only way I can do that is if I have enough time to make my connection, right? In my previous example, I was coming from New York, I was connecting in Moscow and I was going to Armenia. So that plane, that itinerary was actually operated by two different planes and I had a two-hour connection time so I could make that. Now suppose those two different planes, the timing didn't work out and I only had say a 20 minute connection, 30 minute connection. Now what? Now AirFlight can offer a ticket going from New York to Armenia. So you can see that the timings must match up perfectly to create these connection opportunities such that now they can offer itineraries to cities where they don't have nonstop service. AirFlight didn't need to have a nonstop flight going from New York to Armenia to get me from New York to Armenia. They needed to have flights that are timed perfectly with just enough connection time to make sure that I can get on both planes. I'm just talking about from your standpoint. Now what about an airline? So airlines have hundreds of operational constraints that they have to deal with. So curfews, you can't fly whenever you want. Airports in certain cities have limitations, you can't fly between say midnight or 5 a.m. because people are sleeping. Major international airports are a slot constraint meaning that you actually have to have a specific slot at that airport to arrive or depart at that time. Gates, right? You need to have a gate that you can park your plane at. So I could have 10 flights arriving at the same time but five gates and then now I have five planes that are sitting on the runway waiting for a gate to become available. And what's a plane is sitting on the ground, it's not making any money. Planes make money when they're flying in the air. And you share runway, you share an airport, not you don't have an airport to yourself as an airline, you share an airport with other airlines. So the number of arrivals or departures you can have at any given moment is as throttle. It's limited because you share this space. So as you can see the problem is is not so trivial. The thing that you're solving for is complicated. The constraints that you have are complicated. So how do we solve this? You have five minutes. So our approach. Step one, clean the data.