Josh 0:00:13.3: 

Welcome back again for our Anaplan Connect San Jose. We're really excited about our two o'clock session. We have a pretty full room, so I think we'll get started. This session is, Planning at the speed of light how NVIDIA scales supply chain resilience in the AI era. It is my proud pleasure to welcome today's panel. We have Mark Gordon, senior director of supply chain solutions at Anaplan, and Khushboo Murarka, planning and solutions architect at NVIDIA, so please give them a warm round of applause. 

 

Audience 0:00:41.1: 

[Applause] 

 

Josh 0:00:45.8: 

One additional note, as well, we'll have a Q&A session at the end, so please hold your questions until the end. Without further ado, I'll pass it off to Mark. 

 

Mark Gordon 0:00:53.4: 

Thank you Josh. I appreciate that. Welcome, everyone. Nice to be here, and nice to see a full room. Thank you. Can I introduce you in a second? 

 

Khushboo Murarka 0:01:00.0: 

Sure. 

 

Mark Gordon 0:01:01.7: 

My name's Mark Gordon. I lead our supply chain solutions here at Anaplan, so I get to see across all of our customers in all industries doing all kinds of really awesome things. This could be the awesomest. We'll decide at the end. I don't know if that's a word, but we're going to use it today. I'm an engineer, not an English guy. The planning challenges that they have, that we'll talk about today, are immense, they're fast-paced, they're really, really, really hard. So when we talk about supply chain in general, and Anaplan, and all that, we always talk about hard, but I think probably you would agree with me, this brings hard to a whole other scale, which we will dive into. Like we said, feel free to ask questions at the end, if you can, and we're going to leave time to take as much as we can. A couple of things today, because I like to keep it lighter, a little more interactive, although, like I said, questions more to the end, today is Earth Day, just so we all know that. Nods. 

 

Audience 0:02:03.4: 

Yes. 

 

Mark Gordon 0:02:04.0: 

Okay. I did look up two other things, which were I think kind of interesting. Number one, the first National League baseball game was played today in history, in 1876. That was kind of interesting to me, and the Big Mac went on sale for 45 cents in 1967. All things happen today. Now you can add the story that Kay is going to tell you into April 22nd lore, because that is going to trump all of the other things that I just talked about. Now, without further ado, can you tell us about yourself? 

 

Khushboo Murarka 0:02:42.6: 

Sure. Yes, Khushboo Murarka. I also go by Kay, I just like to be called, so feel free to reach out. I work as a plannings and solutions architect in the Anaplan COE team with Jonathan. You met him this morning on the keynote. We have a lot of our COE team here, and we have some of our customers from NVIDIA, so a pleasure to have them. While most of us see the growth of NVIDIA from outside, my job is to look at it from inside, and architecting models which can [?be worth/work 0:03:18.0] $X million datacentre to supply chain models. That's what we do. That's what I do at NVIDIA. 

 

Mark Gordon 0:03:27.9: 

So let's talk about the company for a second, because I think, depending on who you are, you've got maybe a different impression of the company. I'm a little older, a little greyer. I knew the company as graphics cards - gaming, graphics cards - and so on so forth, like that's what the company is. Then, all of a sudden, people are telling me, 'That's not what the company is, Mark. The company is an AI company,' and I'm like, 'No, it's not.' So tell me, how did the company pivot from what I thought it was to what other people think it is, and more importantly, what do you think it is? 

 

Khushboo Murarka 0:04:10.5: 

Great question. So the secret is, we didn't actually pivot from the technology. We pivoted its application. GPU cards, when we launched, it was used and the technology that was in place was for parallel processing, and then about a decade ago, when researchers found that the same technology can be used for AI and deep learning, it is when this whole thing came in the picture. NVIDIA leaned onto that, released the CUDA architecture and the CUDA platform, which made the GPU hardware available for general-purpose computing. For us in the supply chain, what that means is we are not just getting demand from the gaming companies, we get the demand from across industry, so healthcare, auto, and so on. 

 

Mark Gordon 0:05:07.2: 

I don't think it's a secret. The company's grown a lot really fast. So let's dive in a little bit into the planning challenges that exist. So supply chain, like we already said, it's hard. It's hard in any circumstance in any industry. It's even harder in yours, and it's even harder for your company. We talk to clients all the time. I do talk about NVIDIA all the time. I'm like, 'What a wonderful problem to have,' in the sense of demand and supply. It way outstrips, and you've got to make a difference set of choices, and you've got different problems to solve. Then, people are quick to tell me, 'Problems are never wonderful,' like their problems, and they're painful. Tell me a little bit about planning in this phase of growth that you guys have, and how does that all come together? 

 

Khushboo Murarka 0:06:03.9: 

Yes, so planning in this hyper-growth, traditionally, we would go into looking at the historicals. I have my supply chain planner here, but in the traditional system we used to look at the historicals, we would project, and we would go that way, but here, at this company, the demand is exploding. So when demand is exploding, history just comes history, right? It's not just about volume, it's about the volatility and it's about the velocity. So customer signals are changing, product mix is evolving, and our manufacturing process, there are products which need that packaging. The packaging lines are long. So we are planning two to three years out, and for that to happen, we do need models which can support those kinds of scenarios in place. We need that scale of data to be there in the model, and that's what it looks like. 

 

Mark Gordon 0:07:14.1: 

Yes, I like your perspective, and it's always refreshing to talk to you guys, because a lot of times we'll get that. We'll get, 'Hey, I'm growing so fast,' or COVID happened, or there's this disruption or that disruption. History is not going to tell me what to do in my business going forward, which I get, and there's maybe two schools of thought from there. One is, history's not going to help me, so therefore a planning system is not going to help me. Or, like you suggest, history may or may not help me, but we need to have this scenario-based planning tool that's going to help us manage, because somehow we have to get our arms around this and figure it out. So I do completely understand that. There's another wrinkle that you have. In addition to the market, your company, suppliers, all that kind of stuff, there's, I don't know, geopolitics and disruptions, and all that kind of stuff, so now you've got this also added layer of extra disruption on top. So how does that factor in, make life harder, but for you to grasp onto? 

 

Khushboo Murarka 0:08:33.2: 

Yes, I would say, at NVIDIA that's always been the case, tariffs and all of that. So if it were a stable economy, we would be designing for a low-cost model, but since geopolitics is in play, we would make sure that we are designing anything which is resilient. The model should be resilient enough to reroute our supply chains. Let's say we have some politics come in at a different site, so we need to make sure that the planners have the ability to reroute the planning, and the inventory needs to be reallocated to a different site. So all those need to be factored in, which is making sure that all of these scenarios are factored in while building, to make sure we are building a very resilient model. 

 

Mark Gordon 0:09:27.1: 

Yes, makes sense. Let's talk about maybe company culture, company philosophy, as it relates to the way the company performs, as it relates to personnel, resources, and so on and so forth. We heard this thing - I think Jonathan teased it earlier this morning - this notion that we all read about and hear about, the speed of light. 'We operate at the speed of light.' What does that mean? What does that actually mean? Not like the stuff we read in here, but what does that actually mean from the inside? 

 

Khushboo Murarka 0:10:02.6: 

Great question. 'Speed of light,' for us NVIDIAns it's not just a slogan, it's a mindset that each and every team is made to think. So what's the fastest way to get to this, and what's the roadblock for this? Every team in the company is expected to think that way. In the planning world, it actually has a direct impact, so we need to make sure that what we are building for the planners is - every click that we remove, every workflow that we automate, is giving back time to the planners. They need to be not wrestling with the tool. They should be spending time on the actual planning, so that's what it is, and we would make sure we bring the capabilities at their fingertips. 

 

Mark Gordon 0:10:58.4: 

For us reformed former engineers, continuous improvement is a real thing, regardless of what you do, and I think your perspective is always a good one. Like, how do we get even a little bit better? In a lot of ways it's like, how do we get a lot better, but even continuing to turn the screws and make improvements is a huge thing. You're at the company a few years now. 

 

Khushboo Murarka 0:11:22.5: 

I am. It's been a couple of years, yes. 

 

Mark Gordon 0:11:26.5: 

Rewind to day one. I can't imagine - I mean, I can imagine - what is it like when you step into a company, given the planning challenges, the environment, the macro-economics, everything that we just covered, and you show up on day one? 

 

Khushboo Murarka 0:11:47.7: 

When NVIDIA happened I was already in the industry for 15 years, and having worked in the industry for 15 years, it was not just about learning about these new business processes, but it was about immersing yourself in this high-velocity environment, where impact is the only thing people care about. Then, for me it was not just filling the role, but it was also about how can I bring my unique skills and push the frontier forward. So that's the thought. That's where I started. That's my day one. A lot of excitement. Then, what was good is, when we talk about NVIDIA's culture, it reminds me of my days in Mumbai. I come from Mumbai. I don't know if any of you have been to Mumbai. When we go to the platform, you are just thinking about boarding a train, but it's the sheer force of the people that carries you into the train. NVIDIA's culture is like that. The culture's not just supportive, it actually pulls you, and every team is demanding excellence from each other. So even though you feel it's a high-velocity environment, but once you're there, if you have the passion, you just get pulled into the culture. 

 

Mark Gordon 0:13:24.8: 

We were talking about this before, and if you haven't Googled 'images busy train Mumbai' and go on, totally do it. I thought the New York Subway was a big deal, and then I realise it is not. So again, the level of planning challenge is enormous and immense.  

 

Khushboo Murarka 0:13:52.0: 

I want to cover the third piece here. So given that's the case, you're pulled by the culture, but the third thing which I try to do is, I had the day-in-life-of sessions with my planners. That's how you understand what they are going through and what they are building. So my whole idea was not to just be the order-taker, and be the strategic partner, so consulting, advocating them, that, yes, you want this, but let's make sure that this is also taken care of, and so on. 

 

Mark Gordon 0:14:29.0: 

Let's talk about that, because there's a couple of different ways to approach the role, right? It's like, I'm just going to do what I'm told, or I'm going to do whatever the business says, or maybe I'm going to go the other way and not be argumentative, but challenge back, or whatever. So there is a difference in the approach of the role to, like, I'm just going to execute, to I'm going to be a strategic partner. Maybe if I think it through, there is a phase where you start as one and morph into another, or maybe you just start in that latter camp. I don't really know, but what's the key to be seen not as just an executor and order-taker, but to be seen as a strategic partner for the business? 

 

Khushboo Murarka: 0:15:21.5: 

Let's say, if a person is just joining the team, you would need to build the trust, for sure. That's the first and the foremost important thing, and to build the trust, it won't just come as is. You're in the first 15 days or first two weeks, you need to somehow show and get that trust, and the models, I'll just talk through some story. The models that I was working on and have been working on were built, were first launched in 2020. Most of the models, I'd say, in NVIDIA, it's like quick. We would have six to eight weeks, and the model is up. We would have a use case, and based on that the model is up. So the 2020 model, and then I am in - fast-forward 2024 - so it already feels that the model is built in the previous era. Now, it did meet all the requirements then, but then, if we look at NVIDIA's graph, products, demand has been growing exponentially, and geopolitics, as you talked through, so there's lots that is needed. Then, when I go and talk to the planners, they are going back to the spreadsheets.  

 

Mark Gordon 0:16:48.3: 

That's not good. We don't want that. 

 

Khushboo Murarka 0:16:50.2: 

That's not what we want. So we wanted to bring back the trust and bring back them to the model that they were using, so making some quick wins there, accepting that there are some issues going on, giving them a transparency that, okay, we'll go and fix this, but it's going to take this much time. So they have an idea; they understand that we need to just make sure that, for these two days, or for this 24 hours, if we can make do, fine. So that was on the classic side. Then, we were, on the parallel, also thinking about moving to Polaris, because the scale that we were growing from 2020 to 2024 to 2026, it's a lot. So product 40, 80, 120, 200, that's the scale that we are working on, and then we want a lot of scenarios to be in place, we want multiple years of data. So even though we correct, we perfect this classic model, it would be difficult to fit in the new requirements that come in, and so we decided to start this whole Polaris migration, and with the Anaplan team - I don't know if you're here CS and the PS team - thank you for all your support, but we started working on the Polaris migration. So quick wins on the classic side, and then parallelly working on the Polaris.  

 

Mark Gordon 0:18:33.9: 

Let's go off-script for a second. Quick wins, I get that. Can you look back at something and say, 'I made a mistake,' or, 'I wish I would have known or done something different,' whether it was the way you were handing the growth, whether it was the framework and in terms of the way you were operating, or did you know - again, back to the speed of light - we just have to accept, and the company accepts, that we are going to go fast, the market's moving fast, we're going to outgrow and break stuff and have to redo? Was it a conscious thing, or do you think, like, man, if I had to do it again, I would have done something different? 

 

Khushboo Murarka 0:19:14.6: 

Some of it, we would make sure that the Polaris that we are using right now, we make sure that we take our past learning, but then, at that time, when we put in the models 2020, we did what we could that time. We gave them the capability of how we can bring data from different models, and it's a working model. I don't think we had projected the growth. It was exponential. 

 

Mark Gordon 0:19:45.3: 

You couldn't have, really. 

 

Khushboo Murarka 0:19:47.0: 

Yes. Our model size is, I'm not talking about a 200, 300 GB model, I'm talking about 700 GB models. So even if we add a new enhancement or new features or new functionalities in there, there's no way we can accommodate that in a 650 GB or a 700 GB model. So that's what it is. That was one reality, and now we had to move, but now that we have moved we need to make sure that we do not repeat that cycle. 

 

Mark Gordon 0:20:19.7: 

Got it. So I like this, because I think this is a pretty reasonable snapshot for what you've been doing and when, and I look at things like this - I'm sure everyone that looks here probably picks up something different - the thing that I pick up on is, to your point, 2020 to 2026, all of these different hexagons, all these widgets, everything else, the thing that I like about this is the nature of Anaplan and the nature of its flexibility, but working with our customers. This is not a roadmap for anybody else. This is a roadmap for you. We've got lots of other customers creating their own path in here. So I look at this and I see two things. One, it's cool that you've been a customer since 2020, and have stayed with us, and two, that you've built all these things on top of and extending, and essentially, meeting the need with the solution. So when you look at this, can you take us through - maybe it's the broader COE too - how do you evaluate projects, and how do you actually create this and say, 'This is the next thing we have to do. This is the one after that,' and put that on and put some kind of justification for your own time and resources? How does that work? 

 

Khushboo Murarka 0:21:48.1: 

Yes, I'd lean onto Jonathan for a more strategic answer, but to me, if we see this whole graph, we initially started with the basic forecasting. We had the materials planning for boards, chips, all of that. That was diving into the Anaplan quick wins, build, RBP, and stuff, and then as we move forward, if you think about the growth of NVIDIA, that's when, in the mid era we stepped into the liabilities and the country of origin. Those were the quick wins. This is something we got into it because of the mid era, because of the growth, and now that we have all this suite, and there'll be more which will come in, but then we want to make sure, hey, wait a minute, we have all this here, but the model size is growing. Come 2025, 2026, we've thought about it, and we are migrating to Polaris. 

 

Mark Gordon 0:22:55.3: 

Sounds good. So we've covered a few things, and I want to talk a little bit more to round out the perspective. One of the things we talked about was you not just being executor but a strategic partner. The other is creating a roadmap, everything that you just discussed in that last slide. There's got to be some notion to others in the organisation, executives, other stakeholders, like, hey, we're doing this work, it's impactful, we can measure it, it is helping us, so that we can do more. So what is that like within the company? 

 

Khushboo Murarka 0:23:37.9: 

Some of the lessons learned - obviously, with every exercise we'll have some lessons learned - I'd say, now that we have migrated to Polaris, we want to make sure that we are not building for just the current demand or current requirements, and all of that. We want to make sure that the product that we have built is for the future, and it's scalable, and it can take more requirements; we have the geopolitics. So it should be susceptible to a lot of other factors, and it needs to be scalable. We do not want to start with a 100 GB Polaris model already, and we have so much sparsity in the data itself, so if I just talk about the scale, we are at six of TGB went to, say, ten, fifteen GB, and we want to further optimise. So we want to make sure that the model that we have built is scalable, and for all the models, and we want to put some optimisation in place. That's definitely the case, and we do not want to repeat history, so that's definitely the first. The second, I know, everyone, we talk about the AI factor, but I feel, to me, the human factor is extremely important. That is the relationship that a solution architect will have with the planners, or the supply planning team, so to make sure that we are by their side, we understand their pain point, and then we also have put governance in place. If we do not have any governance in place, then the model is definitely not going to be scalable at all. That's for sure. 

 

Mark Gordon 0:25:30.5: 

Was that an already-accepted thing, or was that an unnatural or a thing that you grew into in terms of having the architect sit alongside and understand that? Did the company embrace that? Did you force that on them?  

 

Khushboo Murarka 0:25:46.9: 

No, and the company actually embraced that, and I feel more strongly about it, especially NVIDIA is building things very fast. So we may not be thinking at the same moment about the governance, so now that it's scaling and we've seen the models, we are moving away from the 2020 models, so we want to make sure that this governance is in place, so that we do not repeat it. Governance coming from all the best practices, from the DISCO architecture, and also the whole requirements gathering and the ticket management in place. So we also make sure that we have multiple rounds of testing. It's not just quick build and give it to the business. So we want to do that testing, we want to log all the requests, and then we want to go back to them, work on them on a priority basis, and so governance, for me, is the intermodal and outside the model also. Then, the last piece, very important. Staying close to the platform roadmap has really helped us. So now that we've moved to Polaris and we have been using Polaris, so we want to stick close to the platform as to what is coming. I went through a lot of these features, CoModeler and Insights, and so I would prefer and like to be close to the platform roadmap. Even there are a lot of new features that have come in the management reporting, so how we are using that now instead of one of the older technologies that we had, so we are trying to do that, and we'll try to stick to the plan [inaudible 0:27:41.4]. 

 

Mark Gordon 0:27:42.2: 

Great. All this collective experience, if you had to maybe pick one thing, one piece of advice for people starting their supply chain transformation - it could be anything - I'm trying to wrap my head around it - it could be process, it could be governance, it could be strategic partner, it could be your honeycomb map, start in one place, put metrics, whatever - what would you think it is? 

 

Khushboo Murarka 0:28:13.2: 

I'd say we should respect what we have. I mean the technology, but we should always remember that our goal, if we think about the higher-ups, Polaris or technology is one thing, but what they most care about is, is the tool compatible enough to take care of the scenarios, if we are able to do what we need to do. So sticking to the actual goal of what we are trying to do, and then using the tool as your vehicle, is what I will tell everyone to focus on, and make sure that, again, staying close to the platform roadmap and using that for excellence. 

 

Mark Gordon 0:29:01.3: 

Makes sense. I feel like I... 

 

Khushboo Murarka 0:29:04.4: 

There's one more, actually. We sometimes, while building, being from the COE team or being a solution architect, we would start thinking from a builder's perspective, but we need to be thinking from the planner's perspective, what they actually want, and how we can enable them, and think more from a business side, the functionality side, as opposed to just this will save a space and stop. Yes, those things are important, but the planner's mind and the planner's frame of thinking is more important. 

 

Mark Gordon 0:29:44.6: 

Well, a planner's mind is probably a scary place to be. 

 

Khushboo Murarka 0:29:47.9: 

Yes. [Laughs] 

 

Mark Gordon 0:29:51.5: 

I could talk to you all day about this. We have a packed room. I've got a million more questions, but I also want to open up to the group here, to ask any questions while we can. I'll bring the mic around. Everyone just keep your hands up. 

 

Audience 0:30:10.3: 

Hi. Thank you for the presentation. I'm curious about your experience with the migration to Polaris, and my question might be a little specific. Did it really reduce the size of your model from, let's say, 700 GB to half of that, or what was it like? 

 

Khushboo Murarka 0:30:27.3: 

The reduction is mind-blowing. I'm talking about a 650, 700 GB model reducing to 15 GB. 

 

Audience 0:30:37.0: 

Wow. [Applause] 

 

Khushboo Murarka 0:30:37.9: 

So that means, from a customer perspective, we don't have to pay for that.  

 

Mark Gordon 0:30:43.3: 

That we can talk about separately. 

 

Audience 0:30:47.0: 

One more question. Is Anaplan being used by other functions outside of supply chain? 

 

Khushboo Murarka 0:30:53.2: 

Oh, yes. As of now, our team is mostly focused on the planning side, but yes, we are in conversation with Anaplan to be using it for other functions. 

 

Audience 0:31:06.9: 

Okay. So you're doing quick wins with other function team? 

 

Khushboo Murarka 0:31:11.2: 

No. Actually, my team is mostly focused on the planning, but we [?ARC 0:31:16.1] and we're saying with the other teams ARC and we're staying with Anaplan to get there. My quick wins are within the supply chain planning team itself. 

 

Mark Gordon 0:31:26.4: 

We do know supply chain is the centre of the world. [Laughter] 

 

Audience 0:31:36.0: 

Hi. Thanks for the presentation. It was really good. I have a similar question. We're also making the move to Polaris. We already have the licence, but we still have a lot of models to manage, or modules, I would say, because we have a [unclear word 0:31:48.6] model, we have still many models to manage on the classic workspace, and it's always a headache, because most of our calculation engine is there. So how do you manage that if you still have your models on classic, that work space? Every time we want a report, we do not have enough work space to put together for the end users, so how do you manage that with this lightening of speed? 

 

Khushboo Murarka 0:32:18.8: 

That's where this human factor comes into play, and we faced this issue several times when we are debating between the scale of data that we want to keep, let's three years' worth of data is what we want to keep versus we have eight scenarios that we want to maintain. So sometimes that's a juggle, and we will try to have that transfer and conversation with our customers to see this is the way we can - either we can keep all eight scenarios, or we can keep the three years' worth of data. What is more important to you at this point in time? Then we give them a time that this is going to be the case for the next six months, and then we partner with them and they are very cooperative, so they understand what the model is going through, and everyone understands the sparsity of the data. So we'll get answers from them, and then we'll act accordingly, but I think it's a good idea to maintain your classic model, and that's what was a win in the model that I was working on. We didn't give up. Although our Polaris was a rebuild, what we still wanted to make sure that the classic model is operational, and it's being used. Even to a limited capacity it is being used.  

 

Audience 0:33:48.4: 

You answered my question. I wanted to ask you, did you migrate or did you rebuild from scratch with all [?these sort of things 0:33:51.8]? 

 

Khushboo Murarka 0:33:53.3: 

Yes, we rebuilt from scratch. We did use the framework from the classic. So how we do it, we have a process, but then, at the same time, we actually did it in a phased approach, so this is entire capacity planning of all the manufacturing processes. So since we have the classic model in place, we brought in a few phases, and then one phase after the other of the manufacturing, then we added the sensitivity analysis over it. We also worked with the data integration team, to make sure that the data is switched to a fast-paced integration. 

 

Audience 0:34:35.2: 

Are you using ADO, or not yet? 

 

Khushboo Murarka 0:34:37.9: 

We are about to start using ADO. Yes, we had some POCs done, so we will be switching to ADO pretty soon. 

 

Mark Gordon 0:34:52.5: 

This side of the room [inaudible 0:34:53.7]. 

 

Audience 0:35:01.8: 

So what was your level of [?grain 0:35:02.6] of planning? Do you guys go down to a daily grain, weekly grain, monthly grain, and what's your - you said it's about three years off the horizon. 

 

Khushboo Murarka 0:35:11.8: 

Yes, three years. It depends on the planning model. We have three years. We would have one year to three years. in certain cases we have more than three years. The lowest grain, normally, on an average, it's weekly, but in certain cases we go daily also. 

 

Audience 0:35:30.4: 

It's all built into the same model, or have you got multiple models? 

 

Khushboo Murarka 0:35:33.7: 

So one specific function is built into one model. Like the hexagon that you saw, those were all different functions, or those are all different products. So each model and each product has a separate model in place, and then we have capacity planning, we have build, we have RBP, and we have customer allocation, so all of these different models. 

 

Audience 0:35:59.8: 

How many scenarios do you guys have in scenario planning? 

 

Khushboo Murarka 0:36:04.7: 

I would say we have a bunch. I would say we have a handful of scenarios, and it depends what the business is asking for about that. Then, every model differs in how many scenarios they want to hold onto. In the model that I have been closest, we tried to keep it to eight scenarios. 

 

Audience 0:36:25.9: 

Do you guys have multiple drops and versions of everything that has been planned out? 

 

Khushboo Murarka 0:36:32.4: 

Oh, yes. We have versions all the time. We've automated that quite a bit, so every week we will have new versions come in for, say, demand or throughputs and all of that, and then we also do rollovers in the model to make sure that the current scenario's data has rolled over to the next scenario, and so on. 

 

Audience 0:37:00.9: 

Have you guys used PlanIQ in the past, and are you guys using Forecaster right now as part of your [?bundle 0:37:07.4]? 

 

Khushboo Murarka 0:37:08.2: 

That's a great question. We didn't use PlanIQ, and as of now we may venture into Forecaster. The only reason being that NVIDIA's architecture, it's difficult to just use a custom-build application. We like more flexibility, so we go from the ground up.  

 

Audience 0:37:29.2: 

So was it a lot of statistical modelling built in into the system? 

 

Khushboo Murarka 0:37:33.2: 

Oh, yes. We have that in place. That's where we start.  

 

Mark Gordon 0:37:39.2: 

I think we have a few more minutes. 

 

Audience 0:37:42.4: 

I have one more. In your view, did you see any limitation of the number of reports coming in? 

 

Khushboo Murarka 0:37:51.9: 

For the Polaris? 

 

Audience 0:37:52.6: 

No, for ADO integration? 

 

Khushboo Murarka 0:37:54.9: 

We didn't quite see any limitations. 

 

Audience 0:37:57.1: 

So how much would be the volume in one integration? 

 

Khushboo Murarka 0:38:02.0: 

The POCs that we did were mostly for a certain limited set of data. Now what we'll be embarking on that journey we'll know more.  

 

Audience 0:38:10.1: 

Okay. So we're on the same [over speaking 0:38:11.5]. 

 

Mark Gordon 0:38:11.9: 

Stay tuned, I guess is the answer. 

 

Audience 0:38:14.4: 

Thank you. 

 

Mark Gordon 0:38:15.4: 

We have time for one more question, if anyone's got one.  

 

Audience 0:38:21.8: 

[Unclear words]  

 

Khushboo Murarka 0:38:27.4: 

Yes, it's everything. As of now, it's in one. So all our models are the same for - I mean, it's one model for all the regions. 

 

Mark Gordon 0:38:38.7: 

I lied, we have - that was a quick one, so we can have one quick... 

 

Audience 0:38:44.2: 

How long did it take you to do the forecast, because you have so many scenarios, and weekly, monthly, and you are going even daily, so how you close your forecast? 

 

Khushboo Murarka 0:38:53.7: 

It's a weekly cycle. We just work on it, and the planners work on it. 

 

Mark Gordon 0:39:02.2: 

I love the fact that you're so calm about all of this. What she's saying is not easy, I guarantee it. Thank you so much for sharing your story. 

 

Audience 0:39:15.4: 

[Applause]