Data Analytic Tools and AI: A Winning Combination for Formula E Racing

This transcript has been edited for length and clarity.

Michael Krigsman: Formula E Racing involves cars, speed, data, and advanced technologies such as AI and machine learning. We're live in Brooklyn, New York, at the site of the Formula E Race. Sylvain Filippi is the managing director and chief technology officer of the Envision Virgin Racing Formula E team. Sanjay Srivastava is the chief digital officer of Genpact. Gentlemen, thank you so much for taking time and for being with us on CXOTalk.

Sanjay Srivastava: Thanks for having us.

What is Formula E racing?

Sylvain Filippi: Formula E is the first international racing championship dedicated purely to electric cars. The idea is to achieve two main things: develop technologies that will trickle down to make electrical cars better and also to use the platform—basically, motorsport as a marketing platform—to talk about these technologies and all the things we do around it.

Michael Krigsman: What is your role?

Sylvain Filippi: I am managing director of the race team, so my primary focus, you would say, is to make sure we win races with all the stuff that comes with it, so there's an engineering aspect and so on. Also, of course, responsible for marketing, commercial, and media and making sure that the world knows about it.

Michael Krigsman: Sanjay, tell us about Genpact and tell us about your role.

Sanjay Srivastava: Well, Genpact is a global professional services company. We deliver digital transformation for large, global clients across the world. I'm a chief digital officer, which is to say that our teams help our clients with artificial intelligence, automation, analytics, and experience-based approaches to transforming the business and business processes.

Michael Krigsman: What's the relationship between Genpact and Formula E?

Sanjay Srivastava: We're partnering very closely on helping Envision Virgin Racing with the data, the analytics, and artificial intelligence. It's very similar to the work we do with clients across the globe in helping them win and get ahead in the race.

Sylvain Filippi: To extrapolate on that, you could think that motorsports are just about developing a fast car, put a fast driver in it, and win races. Actually, there is so much more to it. This is only the beginning of the story.

What really makes the difference between winning and losing races is analyzing the data and there is a lot of it, not only to get the good insights but do it in a really timely manner. Formula E is a one-day event, which is very rare in motorsports, which means we have very little time, literally minutes between sessions, to analyze all the data and make the correct decisions.

Basically, the Genpact guys are helping us to make sure we go through all that data, gather the right insights and, hopefully, make the right decisions.

What data is collected and analyzed during a Formula E automobile race?

Sanjay Srivastava: We're helping build a foundation for data across many fronts, right? There is a set of telemetry data that comes right off the cars, and that's a streaming set of large data sets that are coming through as the cars are on the road. There's also data on competitive and driver maneuvers because we can track the rest of the drivers as well, analyze that, and then, over time, machines learn from them.

Then there is data that comes from weather, satellite, traffic, and road patterns that allow us to understand traffic and local conditions. What we do is we interpolate all of that data to try and come to a prediction. We'll talk about that a little bit.

Then the fourth set of data is actually the audience data. You have to have a really solid foundation for data because you're bringing in data from four different sources, it's streaming, and it's large. Once you have the right foundation in place, then you actually enable yourself to be able to drive analytics and AI on top of that. That's really what we're doing here together.

Michael Krigsman: All of this, of course, is real-time.

Sanjay Srivastava: It has to be real-time. Yeah.

Michael Krigsman: This is not, "Well, let's go back to the office, we'll look at this for three or four months, and we'll get back to you."

Sanjay Srivastava: No, it's not. It is real-time. What's even more important is, you've got to get the foundation for data to really be able to even get started. Then you have to drive AI and analytics to be able to make real-time inputs to the decision.

Then the third part of the challenge is, how do you actually expose it? How do you drive the user interface? How do you actually get the users to experience it in a way that they can actually act on it instantaneously? That's been a really interesting part of the discussion, which is getting the experience right so it's absorbable at the right second instantaneously and can be acted upon.

Michael Krigsman: Sylvain, the user, in this case, is the driver.

Sylvain Filippi: Well, it's the drivers and the engineers, mostly, actually. The driver's main job is to drive the car and then for us to help the driver basically make the right decisions.

Back to Sanjay's point, I think what's really interesting here is that we can really use motorsports really as a case study for how to use data because many companies have a lot of data. For sure, we have a lot of it because every mileage on the car generates a huge amount of data. What's interesting is that we have to learn to really structure the data really well so we can then use it really quickly, as we touched on.

If you have years to look at data, it's actually not so difficult. But what companies will need to do more in the future is to use all of their data and use it really quickly. Ultimately, up to real-time. So, this is what we're doing, by definition, because, during a race, we have less than a second to make decisions between using an attack mode, overtaking, or so on. We have to do a lot of homework to make sure the data is really structured and then Genpact is helping us with the correct algorithms and all the things you need in place in the background to make sure you can make these quick decisions.

What are your primary data sources during an electric vehicle race?

Sylvain Filippi: One of the main changes, again, which could be a change for many companies, is the point is that the data comes from very different sources and very different formats. It would be easy if it was one big spreadsheet, right? But you have the telemetry coming live from the car, which is a particle format.

Then we have most of the data we get from the car, which we download when the car comes back in the garage. That's hundreds of channels all at different frequencies and so on.

Then you have the radio from the drivers, from our drivers but also the other teams. Then you have the TV feeds, so video format coming from the broadcast. The list goes on. There are many sensors on the car that all send their own signal.

What's really difficult is that all these come in different formats, so there's not enough time to look at all these things individually and then try to make sense of it. So, a lot of the work we've been doing is to basically develop an infrastructure that integrates all of that data, synchronizes against the time of day, really literally, so the engineers, between sessions or even during sessions, can say, "I want to look at what happened at that moment," and get the whole picture because of the danger.

I guess it's the same for many companies. If you only see part of the data, you could be misled into making the wrong decision. If you look at one part of the data on the car, it could lead you to think that this is happening but, actually, if we have all of the data, then we're like, "No, actually not. This is happening because that, over there, was happening."

Sylvain Filippi: It sounds obvious, but it's really difficult to do. What we've been quite good at doing and getting better every day is to structure all that data to make sure that we have the whole picture really quickly.

Sanjay Srivastava: One of the things I think we've done and it might be a good way to think about this, Michael, is to think about layers of a cake. You're building a cake and you've got layers.

Clearly, the first layer is actually the layer of data which is the foundation. We talked a little bit about all of the challenges of getting it together in the right fashion.

The next layer, if you will, is the AI and the analytics layer. What you're talking about here is, in the end, this is a prediction problem. We are going into a race and we have to predict how many laps are still remaining so we can plan the management of our energy better. It's a prediction problem, if you will. We're using artificial intelligence and AI to be able to bring that out and be able to come back as soon as possible in the race with a good prediction on what's remaining so we can manage the energy better.

The next layer of the cake actually is what I call bilinguals. It's the ability for our teams to work together in a fashion that we combine the understanding of the domain, in this case, racing and all of those parameters around it, with the science of AI and data analytics. That “bilinguality” is important because, at the end of the day, you have to still contextualize the readings. You still have to go orient the machine learning. You still have to distill the analytics that comes out in the context of the race itself. That's a really, really important part.

As I said earlier, the final layer of the cake is actually the experience because now you've got to get it back to someone who is going to act on it in an instant on that piece of advice. How you expose that information becomes very important, so it actually becomes actionable. We've thought about it as different layers of the cake and I think that's one way to approach that.

Sylvain Filippi: That's perfectly correct. What is also interesting is that it's not just our data. It's also the competitors. At the end of the day, we are here on the track to beat all the other guys. Of course, we know what we know is our data, but trying to predict what the other guys are going to be doing is super important and, obviously, difficult. This is what we are starting to do.

Once we have done all of this, we need to also somehow integrate the behavioral patterns will be of the other teams. That should hopefully provide the best outcome.

Sanjay Srivastava: We might want to spend just a minute on what actually happens in the heat of the moment, right? Actually, right at the start of the race, we have to start estimating the number of laps remaining.

Sylvain Filippi: It's a timed race, yeah. Yeah.

Sanjay Srivastava: It's a timed race, so it's not a traditional race in the sense that there's a distance that you have to cover and who gets there fastest. It's a specific amount of time that you have to race for and plus a lap. Then you have to go as far as possible or you have to be first in that.

Key and critical to this is actually estimating how many laps are remaining at any point in the race. The reality is, if you were driving on a straight road at a given speed for a specific distance, you can actually figure all of that out.

In a race, that's not the case. The situation is changing on the ground. There's weather. There's traffic. There are other cars, actually, and then there are moves that other drivers are making in the relative instance itself.

What we have to do is, we have to take all of that data, and then we have to apply analytics and artificial intelligence to predict based on weather, based on track, based on the specifics of the day, based on past behavior, and based on what we expect other drivers to be doing—

Sylvain Filippi: Yeah.

Sanjay Srivastava: –on how that race is going to turn out and then predict.

Sylvain Filippi: Yeah. Yeah, it's a very dynamic position because, to go into a bit of the details, the reason why it's so important to understand the total distance and how many laps remaining is to calculate our energy management. The races are 45 minutes plus one lap. But, as Sanjay said, depending on the pace of the leader, you will cover either 30 laps, 35 laps, whatever. It depends. It could vary for many reasons. It's really critical for us to understand really what will be the exact total number of laps until the end because that allows us to calculate really precisely how much energy we can use per lap.

How do you present electric vehicle racing data to engineers and drivers during the Formula E race?

Sylvain Filippi: We have a variety of dashboards, but what's also interesting, if you look in my garage, I have about 12 engineers that do that during the few practice sessions and during the race. Each of them has very specific roles, so we have developed dashboards for all of them individually. You'll get the race engineers or the main engineer on the radio to the drivers who is really mostly interested in what's happening on the track, what's our delta in terms of timing versus the other cars, what kind of strategy calls we can make, and so on.

Then you have a bunch of performance engineers and these guys are really looking at the car data, really trying to understand how we can make this car go faster. It's a whole set of different data.

Then, in a way, some of the most important guys are the simulation and strategy engineers, also sitting in a garage. They are the people who actually see all of the data, and they are the one who will go on the radio to the race engineer and say, "We should the attack mode here or we should save energy there," and all these things. These have two or three screens, a lot of information displayed, and that's a work in progress.

There are always things we can do better, but developing the model, understanding how we link what the other teams are doing versus where we are is the most difficult part. But we are really quite good and, well, we still have some work to do, always.

Sylvain Filippi: It's a never-ending job.

Sanjay Srivastava: That's right.

Describe the collaboration between the Formula E racing team and Genpact data science experts?

Michael Krigsman: The success of data science, AI, and machine learning efforts always require subject domain expertise, in this case, the race, together with data science and technology expertise around AI. Sometimes, that's a hard collaboration, and so how do you make it work?

Sylvain Filippi: In our case, it's actually not hard at all. I'm a firm believer of, if you want to be the absolute best in whatever field you are, you have to create these collaborations. You have to work with the absolute best experts in their field. That's the only way to be the best.

In our case, it's quite simple. I have a team that is absolutely the best in the world in setting up a racecar, both mechanically in terms of software, coaching the drivers, managing the energy, all the things that we are experts of. I wouldn't expect Sanjay's team to know how to mechanically set up a car. It doesn't make any sense. I have a team of software engineers, but they are nowhere near as focused and experts in their field as Sanjay's team. So, we have to work together.

The beauty, I guess we're a bit lucky in our field where software and data, in a way, even though my team and Sanjay's team do different things, they speak the same language. At the end of the day, it's software, data, and AI. They use the same coding platforms and so on. We're also even lucky that you don't even have to be at the same place. You can be at other different parts of the world. You exchange codes and you do video conferencing. It works fine.

I guess it could be maybe a bit more difficult for an industrial company or something like that. In our case, it's fine.

Sanjay Srivastava: We serve many clients across the globe. If I had to just sort of look back in the work we're doing together, Sylvain, and talk about what are some of the learnings that might apply more broadly, clearly getting the talent and the team composition is a really important part. This notion of bilinguality, which is, you have a team that understands the data science, the analytics, and the AI, but also a team that actually understands the specifics of the race and the conditions around that. I think that bilinguality and that composition is right. I think we've got the talent piece right in this case.

I might just add to that. there are a couple of other learnings that come through, as I think about it. Clearly, you have to enable the best teams with the right data foundation. I think the effort we put upfront on actually getting that foundation right and getting all of the data streams in place in a way that that was usable is really important. As I look across the broader set of clients we serve, I think that's a learning and an opportunity that I think we can leverage a little bit more in getting that foundation right. I look at the industrial world and the reality is, 90% of the data infrastructure is going to be re-platformed in the next 4 or 5 years, and so there's a real opportunity to get this right to set them up for success in the long-run.

The third thing I would add to that, and this is actually the fun part, I mean our guys love to work with Sylvain's team. One of the reasons behind that is a small, little thing, but it's really important. IT's about purpose. What we are finding here, as I talk to my teams, I think what I'm finding is people are looking at this and saying, "Yeah, we're doing all the work. Yes, we want to win the race tomorrow, and that's part of it. But what we're really working on is actually setting up the foundation for mobility in the future. We're going to make autonomous driving safer. We're going to make this more practical, that more people will be able to use that.

I think one of the learnings, at least for me, has been, as we work on projects, finding that purpose and then being able to get that through the teams so this becomes purpose-driven. In this case, we're talking about the future of sort of humanity and how we use e-transportation. That's just been a really interesting insight as we work together.

Michael Krigsman: As we finish up, let me ask each of you for advice to businesspeople on that collaboration, on building the right type of team.

Sanjay Srivastava: I think the advice is very straightforward, and I would say that it's important to think about this in three or four steps. I think getting that data foundation right is super important, particularly in the business world.

Look, the reality is, there are two kinds of companies out there. There are companies whose core value proposition is data and the value the data brings in, and they are corporations that have a value proposition that is more traditional. I think that's one.

I think the second one is this bilinguality, which is getting the right composition of the teams, getting that right, and putting that into place. Then the third one is experience. We talked about making sure that the insights are actionable and we could do something about it.

Sylvain Filippi: Yeah, you're absolutely right. The only thing I will add is that, for me, I always go back to that. I always say, start from the end. What is it that you're trying to achieve? In our case, it's easy: winning races and develop better technology. It's applicable across many industries.

Then most companies would be amazed how much insight they can generate from their data regardless of the business to then achieve that goal. You mentioned the last time, 90% of the data in the world is not used, or something crazy like this. I think our team is a good case study and a good example of how using every single bit of data we get is helping us achieve our goal. It sounds simple, but many people don't do it yet.

Michael Krigsman: Gentlemen, thank you so much.

Sanjay Srivastava: Thank you.

Sylvain Filippi: Pleasure.

Michael Krigsman: Good luck with the race.

Sylvain Filippi: Thank you. We always need a bit of luck!