Erwin Paes - CTO at Optessa

(Joel Beasley at 00:00:00) Hello, my friends. Today we are talking to Erwin, the CTO at Optessa, and we discussed the intricacies of planning and scheduling in manufacturing, how Optessa's greatest strength is their people, and the near future of automated car manufacturing facilities. All of this right here, right now on the Modern CTO podcast. Here we go. This is the Modern CTO podcast.

(Joel Beasley at 00:00:35) Erwin.

(Erwin at 00:00:36) Hi.

(Joel Beasley at 00:00:37) Hey, buddy. How's it going?

(Erwin at 00:00:39) Not too bad. How about you?

(Joel Beasley at 00:00:41) It's going good. I am so excited. Thank you so much for doing this.

(Erwin at 00:00:45) Oh, thank you. Thank you for having me.

(Joel Beasley at 00:00:47) I want to know what life is like in India right now. We know what life is like in America here with the coronavirus stuff. What's going on in India?

(Erwin at 00:00:57) Well, of course, if you're looking at the virus, in the summer—it's not the summer, during the monsoons—that's typically when the viruses here are more active. But that's the time when we had a pretty bad spike out here. But now, actually, at least in Goa, here right now I'm in Goa. I don't know if you know where Goa is, but Goa is a small state on the Western coast of India. It's around 600 kilometers below Mumbai. Typically, it's known as a tourist hotspot because a lot of nice sandy beaches and stuff. So right now, we are full of tourists out here. I mean, typically, otherwise, we have a lot of foreign tourists coming too. But in these days, because of the virus and stuff, we don't have as many. But the Indian tourists are coming here like as if nothing ever happened. So I don't really know. The virus really impacted us in that sense. But actually, it was pretty bad. I mean, a few months ago, we had some known people and stuff who actually either had to be hospitalized or got a pretty serious attack of the virus, or a couple of them also who unfortunately passed away. So, yeah, in that sense. But other than that, it's actually very warm out here, or I guess you would call it hot because plus 30 degrees Celsius, I guess, is something you won't be having nowadays, especially in North America, basically. I mean, you probably have your winter out there, there's snow and stuff. So yeah.

(Joel Beasley at 00:02:37) Well, I actually live in a town that's very similar. I just looked the map for Goa up, and I live in a touristy beach town that's pretty much hot all year round.

(Erwin at 00:02:51) Yeah. No, that's cool. Because the thing is that, like, a lot of people, when I'm having my meetings with my colleagues in Edmonton—I mean, Edmonton sees temperatures like minus 30, minus 25 degrees Celsius, right? So that's the time when I make fun of them, you know, saying, hey, I'm on the beach here in my beachwear. So that's it.

(Joel Beasley at 00:03:13) Just yesterday on a Zoom call, I took my phone outside because half of our team is in Nebraska and it's freezing, snowing, blizzarding, and we have blue skies and it's like 80 degrees.

(Erwin at 00:03:24) Exactly. Exactly. Yeah. Yep. Yeah.

(Joel Beasley at 00:03:26) Everyone hates us right now.

(Erwin at 00:03:28) I know. I know. I know.

(Joel Beasley at 00:03:30) Let's talk about you. How did you get started? Like, did you grow up in Goa? Did you move there when you were a kid? Tell me about that.

(Erwin at 00:03:36) Yeah. I actually was born here. I grew up here. I did my college here. I mean, I went—I was doing my engineering, mechanical engineering. And so as I was doing my engineering, that's the final year wherein I decided that I should—I took some courses in industrial engineering, operations research, basically. So that's the time was the spark, basically, to look more than just engineering. I mean, in the sense, I just enjoyed what I learned in those courses. Like, you know, operations research basically was very interesting for me. And so that led to this journey, basically, which means the journey wherein I was in Goa. From Goa, I went to Mumbai. And I can give you more details later, but just in short, from Mumbai, then ended up in Edmonton, Canada, of course. I mean, that's another story there also, how I actually ended up in Canada. But I can get to that later. And then from Edmonton in Canada, I was there with Optessa for like, now I'm there with them for 15 years. And recently, I spoke with our CEO and stuff that I would like to come back to Goa, I mean, for two reasons. One was, of course, more for family reasons. I mean, my whole family is here. My family, my wife's side family. So it's only me and my wife and my two kids who are in Canada. So they were missing their grandparents and stuff. So that was one reason. And second reason also, we had an India office, and that needed to grow. So it was a—I can do two things. So that's the reason why I have, actually, in the last four years, actually been—I think for almost five now—working from our India office here in Goa, actually. So that's how, that's the story of how I started from Goa, went to Canada, and then come back now to Goa, basically. Full circle.

(Joel Beasley at 00:05:39) Smart move, though. You wanted to be back there. So you analyzed the business. You found, you put together a business case to see if it was possible, if it did make sense, and then you presented it. I mean, so many people have a hard time with the creativity aspect that anything is possible.

(Erwin at 00:05:57) No, that's true. Like, I mean, sometimes, you know, it's also luck and things which fall into place, I mean, in that sense. So I'm glad to be thankful for that also. I have to also be thankful for our late CEO, I mean, Vasudev Kranti. I mean, may his soul rest in peace. I mean, he was a wonderful man. And he actually allowed me this opportunity, actually. And when I put forward the case also, at that time, we were only around one or two people in the India office. Right now, we are around 10 of them. And the only reason, of course, we have not gone more is because, of course, this year, we were not sure what we wanted to do because of this whole virus thing. Right? We're not sure what the impact is going to be in that sense. So, but yeah. But, like, I mean, the office in the last four years, we have gone from two to 10. And actually, we want to grow it further because we're getting a lot of nice, good people here also. So that is a key thing in a way.

(Joel Beasley at 00:06:52) So what does Optessa do? Like, what's the 10,000-foot overview?

(Erwin at 00:06:56) If you look at the tagline, it says optimal intelligent scheduling. Right? But, really, what we try to do is we try to help customers, which are mainly the manufacturing domain customers, to do better planning and scheduling. Okay? So what happens is that typically, people are so much aware of, say, supply chain and, you know, ERP and MRP systems. But many times, the whole idea of a plan and a schedule required to run the plant more efficiently is typically people do not look into it much. But the thing is, they're pretty difficult problems. They're very tough problems to solve. For example, like, I mean, you might have heard a traveling salesman problem. A lot of people use it. Everyone uses it as an example. So if you have, say, like, given a 60-city problem where a salesman has to make a tour of all those 60 cities and come back in the shortest possible time or distance, if you look at all the possible tours which are possible, you will be exceeding the number of electrons in the universe, basically. And so, which means if you actually have a computer to solve and evaluate each and every solution, possible solution, that's going to take years. Right? And most of these problems, what we call the sequencing problems, they have this type of complexity, what we call them the NP complex or the NP-hard complexity. Right? And so these are really the things of what we help our customers. Right? So we say, you know what? You've got these tough problems. We know it is difficult for you. We can help you. Right? We can help you, and there are measurable benefits which you can get, basically. For example, in some of our customers, we have actually shown a 10% improvement in the throughput. Now 10%, someone will say, oh, it's only 10%. But actually, if you go and measure it in terms of the dollar amount, it can be millions of dollars. And one of our customers actually came back and said it was a mid-sized plant, but it was close to like $4 million a year that they have been saving just because they have been using Optessa, basically. Right? So in short, I mean, you can get a lot of benefits by, you know, coming up with the right sequence or the right plan, because most of the times, these days, even though we have so much computing power available, people are still using spreadsheets, still using manual ways of coming up with their plans and sequences. A human being, for that matter, cannot look at more than six, probably six constraints at a time. But when we go in here, when people see the tool and say, oh, okay. You can do this. Oh, you can do that. We end up with like 50, hundreds, 500 types of constraints, and the software does it. Right? So that's the cool part, basically. We are there to help our customers save money if you want one in one line, basically.

(Joel Beasley at 00:10:17) That's a great business because that's an easy business to sell. Like, when you're selling that, you're saying, okay. You're doing this manufacturing process. We're going to watch throughout the entire process. We're going to make that more efficient. And now you're going to save 10% or whatever the percentages for them ends up being, and then they equate that back to dollars. So it's a pretty simple input-output investment opportunity.

(Erwin at 00:10:40) Definitely. For those savvy customers who have been able to actually take these numbers and come up with the actual cost, they have seen the value of it. Right? The problem a lot of our customers or some of the customers face is they're not able to sometimes put in the dollar value. I'll give you a simple example. Recently, we had one of our customers, a big auto OEM customer. They were struggling to find what is the actual benefit if they went with Optessa. What would be the ROI? And what we told them, you know, I mean, one simple way, just count what you can count easily. So, let's say the paint batch size. Right? I mean, in manufacturing cars, the number of paint changes you do has a cost. So every time you change the color in the paint line, there is a cost. There's a cost because you need to clean the line. There are emissions. Emissions have to be controlled. So all these are cost. And in this case, they can easily calculate this cost. They are known. And just look at one constraint. And what they said is maybe one of these paint changes would be like $13 or $15 or whatever it is. So they had an average batch size of paint batch size of two, and we showed them you can easily go more than six. Over the year, that turns out to be close to a million dollars, actually, in a way. Right? And that's just one constraint, paint batches. Now you look at so many other constraints which you're looking at. For example, what happens on a line also. I mean, if you ever seen an assembly line, you'll see sometimes, you know, there is one of the workers who actually sits inside the vehicle and is working on it as the car moves down the line. At some point, then he has to stop because someone else has to get in because his work starts there. So what happens is you cannot have some of these vehicles, what we call the difficult vehicles, right? To have them, a lot of them, back to back because that will stop the line. Right? So every time you stop the line, you probably are losing production of one car or two cars a day. And you put them over the year, that's a huge loss. Right? And that's what I say when you start around 10% throughput. Now you put all these things together. It's much, much more than the $1 million. It can go into, you know, tens of millions of dollars, actually. But the problem they face is they cannot measure this cost of, you know, okay. What if three of these difficult cars came back to back? How much would that loss be? What if four of these cars came back to back? Those difficult cars came back to back. And so that's why, you know, sometimes they find that difficult to do that. But, like, some of the savvy customers are actually doing it, and they're seeing the benefit of it. And especially the big auto OEMs, for example, they have seen the value. And that's one of the reasons why almost we have five of the top auto OEMs as our customers, basically. It's because they've seen the value of what we can give them. And it's been from like 10 years. It's not that it's just recent or something. Right? And that's the beauty of that, basically. So—

(Joel Beasley at 00:13:56) Have you called up Elon yet? He manufactures rockets. He manufactures cars. He manufactures batteries. Have you called him up?

(Erwin at 00:14:04) Yeah. At some point, actually, we had, I think, tried an engagement with them. And I think at that point, they were not too interested at that point in looking at something like this. But at least the traditional—I mean, I know I know everyone wants to move into EVs now. We have got a couple of, we got one of the EV manufacturers also recently as a customer, basically. But, yeah. Unfortunately, at that time, we had gone there and we tried to, like, you know, and they were not that interested at that point, but unfortunately. But I would say it is interesting for sure.

(Joel Beasley at 00:14:41) I was thinking, too, because, like, I have a couple of friends that have robotics companies and things like that. And one of the questions that was running through my mind as I was looking at your logos and stuff, is this something that happens with older, more mature companies when they have some time to sit back and analyze? Is that more of your customers, like the Fortune 500 ones that have been around for a while and are just looking for small improvements versus the ones that are, like, you know, 10 years ago, Tesla was just trying to survive. Right? They were barely making it. And some of these new companies, they're just trying to basically figure out how to meet market demand or get into the market correctly, sitting back and making a 10% throughput optimization. You know, so do you notice that more of your companies are these larger—

(Erwin at 00:15:28) Yes. It's that's what has happened is, actually, more of our current customers are the larger companies, basically. And then also there is another reason what also happens. Right? Most of these companies, they tend to have a cycle, right, and we need to also be able to get in there at that time. Like, you know, they have—because all of these are IT systems. So they have a cycle of, okay, maybe 10 years or something where the time they'll say, okay. Now we need to replace this. We have this no longer can meet our needs. Right? So that's one of the things. So most of the customers which we have, we also were able to get in there at the time wherein their current system could not meet their needs. For example, right now, what happens is one of the two things you'll notice in the market are these two-tone vehicles or multi-tone vehicles. Right? You have multiple colors, tones, on the vehicle. Now if you notice, you may say, oh, what's the big deal of that? But actually, it's a pretty big deal for these paint shops, basically. What happens is if the car goes into the paint shop, it has to wait there for the first tone to dry, and then after some time, has to come back in and get the second tone painted. So now what happens is your paint's body shop sequence is different.

(Erwin at 00:16:47) Your final trim body shop, our trim shop sequence is different. And so now they're finding it that the current system cannot handle that because what they were getting before was a single sequence. And now we are telling them, you know what? We will do all this whole thing optimized together. That means we will make sure that your paint constraints are taken care along with your paint sequence, and your body sequence will be taken care with your body constraints, and your trim constraints will be taken care with the trim constraints.

(Erwin at 00:17:18) And all this will be globally optimized together. No heuristics or in the sense heuristics. And that's what I mean by your greedy heuristics or your typical sort based heuristics or whatever it is. Right? We're trying to find you a global optimum, which means we're trying to give you the best solution possible.

(Erwin at 00:17:36) So that's what we're trying. And like, you know, our customer has found that the benefits they're really getting are huge, really huge, basically, in that sense.

(Joel Beasley at 00:17:48) Is your company like, do you do this with your company in a sense? Like, with the sales process and the cut, do you analyze and mark constraints within your own organization? Do you have the most efficient organization on the planet?

(Erwin at 00:18:03) Well, sometimes we do feel we should use our own product on our own planning and scheduling, you know, but I guess to be honest, you know, every company has its share of firefighting and stuff and things like that. So yeah, I would not say that. I don't, yeah. But what I would say is we do have a really good set of people. That makes a big difference.

(Erwin at 00:18:23) I mean—

(Joel Beasley at 00:18:24) It makes all the difference.

(Erwin at 00:18:25) Yes. Yes. It is. I mean, and I say good set of people, not only in terms of how smart and how good and how brainy or how intelligent. I mean, that is a given, of course.

(Erwin at 00:18:33) But it's the attitude of our people who come. Right? And that's the reason why I have been there with this company for fifteen years. Right? So also a lot of my colleagues are there for fifteen years.

(Erwin at 00:18:45) I mean, people, our turnover rate is so small. Like, I mean, people come and stick with it, and they say they somehow love being in this company. No. Yeah. We're not big like Apple or Google or whatever it is. Right? We're small in that sense. But it's just that the people come and stay with us. And that I think is—it also starts right from the top from our CEO, our late founder and our founder, current CEO, Ashok also. It's the attitude which they have, which is one is of respect for everyone.

(Erwin at 00:19:17) Everyone is respected. Okay. Everyone's opinions count, basically. And everyone helps one another. The atmosphere is very nice and lively and very jovial.

(Erwin at 00:19:30) And it also reflects with our dealings with our customers. For example, each and every customer of ours has said that you are the best guys to work with because, you know, as far as support is concerned. Right? Like, what they say is, like, you know, you guys give us a support which no one else gives. Okay.

(Erwin at 00:19:50) One reason is because we do our own support. Like, we don't have a call center and things like that. We don't outsource our support, basically. Many times, I myself have done support for some of our customers, basically, in the sense because if I find there is some issue because most of the support emails, I also see them. Right?

(Erwin at 00:20:07) So if I see that, okay, there is something which I can contribute because of my experience, I will go and, you know, put a statement in there or something to help out and say, you know what? Okay. Maybe tell this customer try this stuff or try that stuff, basically. So what happens is our own people who work, they go that extra mile for our customers. Just to give an example, just it was just last week, actually, last week.

(Erwin at 00:20:36) What happened was this is one customer. Typically, they have their support agreement between, say, their day shift for their daytime. And they had to get this sequence out because their MRP process was waiting on it. So this thing had to get done. And they send us an email around fifteen minutes before closing and saying, oh, please, please, please, please do not stop your shift.

(Erwin at 00:21:03) Right? I mean, we need to get this thing out. We need to get this thing out. Now the thing is, right, I mean, I was feeling bad of what to do, and I didn't want to force our people to say, no, no. You must stay there because the agreement doesn't require us to do that. But they themselves came and, like, you know, and stayed up. The boys actually ended up staying till around 3 AM. Okay? That is from 6 PM, the previous day, to 3 AM the next day.

(Erwin at 00:21:33) And then from 3 AM to 4 AM, they were now working with the customer to make sure that, you know, the sequence goes through so that on their side, it gets done. Now I did not ask them. I did not say anything to them. They did it out of their own accord, basically. And that was, that is the key. That's the beauty. Right? So it's this relationship what we have with our customers. You know, that we know, we feel, they feel that we are not there just to make money. Right?

(Erwin at 00:21:58) I mean, we are there genuinely there to help them, to make things better for them. And I think that's what basically is the key of our organization, really. We do have better software. We do do cool things, but, you know, that makes a big impact to their customers.

(Joel Beasley at 00:22:15) Yeah. Sounds like you're hungry to solve problems. Right? Because I know what you're talking about, staying up late even though it's—you, when you get hooked into that problem or you're getting so close. Right. Great people tend to just want to see it through.

(Joel Beasley at 00:22:29) You know? Obviously, it can't be like an everyday thing, but when it comes up, I know those moments. Those are some of the best moments, and you remember them too.

(Erwin at 00:22:37) Oh, definitely. Definitely. Right. I mean, like, I remember right from college. Right? I mean, when you do those some of those things where for most of the time, your coursework, you just, you know, okay, forget it. But there'll be some challenging problems which you actually, like, you know, stay overnight or something, like, you know, the whole night through just to make sure that you've understood that concept or you like that with the challenge you wanted to, like, you know, do that. And that's exactly what Optessa wants to do is to solve challenging problems. Right?

(Erwin at 00:23:05) Sometimes, you know, when Ashok and I, we speak, we say, you know what? Really, you know, I don't want to be a CEO. I don't want to be a CTO. We just want to go and solve problems. Maybe someone should come and do that job first, you know, because that's what we love doing, really, in that sense.

(Erwin at 00:23:19) We love doing these problems. Right? And that's one of the things also. We want to, you know, keep going as a company, as a product company. That's that was the key thing. Right? That we are a product company, that we need to keep going, growing this product. Right? Otherwise, if you do not, you stagnate. And then that's the end of you, basically.

(Erwin at 00:23:40) Right? That was one of the reasons I joined Optessa, basically, is because I wanted to be part of a product company. I mean, if I, my, I actually started off in with one of our IT services company here in India. Now it is part of TCS. If you have heard TCS. Right? You definitely heard of TCS. That time, there was this company called Tata Unisys. I'm pretty sure Unisys is still around.

(Erwin at 00:24:08) It was Tata Unisys. There was a joint venture between Tata and Unisys. And I signed up with them. And, you know, being in a services company, one of the challenges was this long sales cycles. I mean, we still have long sales cycles here, but, you know, there, it was even worse. Because, you know, being a services company, one of the problems we faced then was, okay.

(Erwin at 00:24:29) You are trying to come and, you know, solve this big problem. For example, we were looking at an airline crew scheduling problem. Right? And they said, okay. Where have you solved this before?

(Erwin at 00:24:41) And that was the problem right there. Right? We, we have this prototype here. We can show you, and we can do this. Well but, you know, we would like to have someone who has done this thing before.

(Erwin at 00:24:51) And so that was always this constant cycle. You show something as well. Have you done it before? It was chicken and egg. Right? Unless someone gives you the first one, you don't know. You can't, you can't say you've solved this one before. Right? So and it was a constant problem, which, you know, which was getting to me, basically. And that's when I said, you know what?

(Erwin at 00:25:08) I need to get into a product company. I need to get into a product company that innovates, basically, which is constantly innovating. Right? And that is how I actually ended up in Optessa. And, you know, that's why I'm I'm still here. I'm loving it here, basically.

(Joel Beasley at 00:25:22) Are you guys using any AI or machine learning in your product?

(Erwin at 00:25:26) Yes. So if you actually, I mean, what we do is we have a mix of various algorithms. Okay? So there's a mixture of AI. There is mixture of operations research. Right? Mixture of sometimes could be even heuristics, which we use and stuff. The reason is because, as I said, just sometime back when I spoke about the complexity of the problem. Right? I mean, there is no one size fits all, to be honest.

(Erwin at 00:25:55) Because if there was, there would have been this one company who has it, and we would not have been in business. Right? Because they would have got all the business. So what happens is one of our strengths is that because we have a unique understanding both on the domain. Like, for example, I said, as I came from mechanical engineering and industrial engineering, so I have a good understanding of the manufacturing domain. Plus, since I've done my master's in industrial engineering operations research, I have an idea of how we can apply these techniques to this domain here.

(Erwin at 00:26:28) So a user does not need to know what is there behind the scenes. Right? He doesn't need to know and see the complexity of this, what happens behind the scene. So what he sees is there's something in his language or in what he understands, his constraints, his rules. Like, for example, he'll say, I don't want a red after a white because you can get a pink car.

(Erwin at 00:26:50) Example. Right? Now what that translates to behind the scene is something, some mathematical stuff or something else that is for us to figure out and do it. So that way, that's what we do as a product. So what happens is depending on the problem, the approach we use can be different.

(Erwin at 00:27:09) For example, when we are doing, say, sequencing, we will be using, let's say, a simulated annealing based type of algorithm. Now SA was one of the algorithms people use for AI-based optimization, for example. Right? In that sense, if you go back to the times when, you know, doing optimization using AI, SA was used. Genetic algorithms is another one, but it is also used in that sense.

(Erwin at 00:27:33) And there are many more nowadays. You'll find swarm intelligence and ant colony optimization and all sorts of so many names, basically. But they all come under the same category of AI-based algorithms or metaheuristics or evolutionary algorithms. So numerous names are given to them, basically. At the same time, we also use a lot of, say, operations research-based techniques also. Right?

(Erwin at 00:27:57) For example, when we are solving large planning type problems, the underlying problem sometimes is more suitable for using a different type of algorithm, basically. And so that's the way we have, we do have a library of constraint type here, library of algorithms, which we use based upon what the problem may be. And that's the reason we are successful. Because otherwise, what happens is everyone has this one, like, say, one heuristic to say, this is the way you do it.

(Erwin at 00:28:26) You sort your things this way, and that's the way you sequence it or that's the way you plan it. And what happens with that is you, you get typically, initially, a very good, like, your first day sequence will be good, but your second day sequence will be horrible because all your stuff which you could not manage, you're dumping it at the end. Right? It's like sweeping all your dirt under the carpet, basically. And that does not work because now when the time comes to actually send those vehicles to the plant floor, the plant floor basically comes back and says, what is this sequence? I cannot build this because, you know, my plant will stop with the sequence you've given me.

(Erwin at 00:28:58) So that is the thing. Right? So we have AI. We have also operations research-based type of stuff.

(Erwin at 00:29:12) And we are also always looking out, like, you know, for new things which may be coming, which may be useful for us. You know, that would help people. Right? I mean, could be either software or hardware. So many things go into play for these type of solutions, basically. So—

(Joel Beasley at 00:29:28) Are you geeking out about anything in the AI or ML space right now?

(Erwin at 00:29:33) Right. So one of the things to be what we are trying to do in especially with respect to AI and ML is something like this. Right? Is when we give out, say, a plan or the sequence, people come and say, okay. Especially when you're doing a plan.

(Erwin at 00:29:50) These 20 or, say, 200 orders, we could not allocate or they could not be planned. Right? They were unplanned. And the customer or the user comes and asks us, why was this not able to be planned, basically? Right?

(Erwin at 00:30:05) And most of the time, it's not an easy answer. Okay? Now I can give you a little long explanation here just for explanation's sake. For example, what can happen is if someone does a very rudimentary type of an algorithm, which is like a simple search or a simple greedy heuristic, what we call it, we'll say I place this, then I place that, then I place the third one. And, okay, the fourth one, I cannot place.

(Erwin at 00:30:33) So that's the reason I have to stop because of the third one. But in reality, it's just not because of one thing which is there. It's a mix of five, six, or even 10 different things sometimes which are together, which are causing that one thing not to get scheduled. I'll give you one example. You know, just recently when we got this one automotive customer, you know, typically, what they ask us to do is that at the, before we complete the sale, they ask us, here. This is our data. Can you show us what we can do? Okay.

(Erwin at 00:31:02) So we take in our data. We do what we call as a proof of concept, basically, and we show what we can do. And so they gave us two datasets. The first datasets, we actually gave them a paint batch of around 11 or 12 colors, 11 or 12 of the same color. They were so happy. Then they said, you know what? You know? Okay. Okay. This was a different one. Let's take this dataset, which was, like, a one month later.

(Erwin at 00:31:34) And there, we were just struggling to even meet six, get six. And for me, that was very confusing. I said, what is it? Why is it you're not able to get six? And so, actually, what we did is, team, we tried to do some of these ML techniques here on this dataset, tried to do some data mining and analytics and stuff. And what we found was something very simple.

(Erwin at 00:31:51) I mean, this is a simple example I'm giving just to understand. Right? But what they were trying to do is one of the important constraint was they wanted to space out the hybrids. And hybrid is what I said before, like, is one of their difficult cars, for example. So you don't want the hybrids to come back to back, basically.

(Erwin at 00:32:11) They wanted to say that between two hybrids, they wanted a space of 14 other non-hybrids. Okay? And what, what we did was try, what we made that whenever we try to space out the hybrids, our paint batches started going for a toss, basically, completely. And so I asked them, is this guy gonna just try and see, do some analysis, try some of the ML techniques on this and see what he can come up with. And rightly enough, what they found was—

(Erwin at 00:32:43) 92% of the hybrids were of white color, okay? And around 88% of all whites were hybrids or something like that. So there was such a strong correlation between these two features that you can understand, right?

(Erwin at 00:33:02) The moment you try to space out the hybrid, it was spacing out the color. So there's no way you could get your color batches with this constraint in. And so this is what we went back to the customer and said, "Hey, you know what? On this dataset, there's no way. It's mathematically infeasible for you to get your run length of six." And we showed them some calculations, basically.

(Erwin at 00:33:23) And, you know, you either have to go back to your paint shop or to your hybrid—to whoever it is, whether it's a body shop or whether it's your trim shop where the hybrid is important—and talk to them and see, okay, what do you need to do? If you want the best of both, something has to give, basically. Right? So these things are important for us, and the customers are now looking for this type of input, basically, because no longer—okay, now you've given me a much better solution, but I still want to extract more. What is it that has caused you to not even go further? Right? So these are the type of things now that we are looking into, basically.

(Erwin at 00:33:57) And that's not easy some of the times because just today, we had a call, an internal call, and where we found that when you start looking at these possible combinations and things, you end up into some millions and millions of combinations. Right? So how do you actually go through this and make sense out of these and be able to provide them to the user which is useful for them? So that is like the challenge that right now—that is what is our goal right now, to get into those type of things, basically.

(Joel Beasley at 00:34:29) That's interesting information for the company because there's other options than just adjusting the paint or the ratio. You could adjust your business model slightly and maybe hire some people for the—so like, that are specialized in the hard cars so that you could batch them at a more dense rate.

(Erwin at 00:34:50) Exactly. Yeah. All sorts of options, right? Basically.

(Erwin at 00:34:53) I mean, that allows the our users to do a lot of what-if analysis, you know what I'm saying? Another example was, some years ago, one of our customers, right, they wanted to introduce a new model on the line. They already had two models, introducing a third model. Right? Now, typically, once you start introducing more models or more variants, the balancing of the line becomes very difficult. Sequencing becomes very difficult. There are too many constraints to manage. And, you know, one of their solutions was to actually do capital investment and make another line, put another line up. Now you know how much would that cost, right? That would be billions of dollars or whatever. A huge amount, basically.

(Erwin at 00:35:31) So they came to us and said, "Okay, you know, this is what we are trying to do." Said, "Okay, why don't we try it out? Why don't we try it out in our software?" And say, "Okay, if you add these lines, you give me the set of types of vehicles you're going to do, put in your the mix of the vehicles which you're gonna have of these three models, and let's see whether we can actually meet it or not."

(Erwin at 00:35:54) And true enough, we said, "You know what? You can actually meet your constraints and all these various constraints you have, the third model, and you don't need to put this huge capital investment. You have—in fact, you don't need any investment at all." I mean, of course, you need some investment on your current line to rejig it to be able to handle a third model, but you don't need to add a whole new plant, a whole new line, basically. Right?

(Erwin at 00:36:17) So that itself saved the company tons and tons and tons of millions of dollars, basically, in that sense. Right? So those type of things is what I would say is the power of the tool. Right? I mean, when you have these type of systems, you know, people can use them in ways much more than just coming up with a sequence because I need to get out of sequence here because my MRP or my parts order purchasing needs it or whatever it is. There are so many things you can do with it and in so many places you can save money, basically.

(Joel Beasley at 00:36:50) Have you gotten to go visit some of these manufacturing plants?

(Erwin at 00:36:54) Yes. Yes. Yes.

(Joel Beasley at 00:36:55) What's some of the coolest stuff you've seen?

(Erwin at 00:36:57) Oh, yeah. Like, I mean, if you go over the over the years, right, I mean, the automation which is happening now in this stuff, right? I mean, years ago when I was doing my engineering, that's the time I used to go for this industrial tour, just to understand what industry is. And that's the time when I had visited some of those lines. A lot of stuff was manual. Right? I mean, you would find a lot of workers on the line and stuff.

(Erwin at 00:37:23) Nowadays, everything is done by robots and so many cool things. Right? I mean, the number of people on the line are less. A lot of automated guided vehicles, a lot of, you know, those type of things. And, you know, which is basically that people are moving to more and more automation. Right? So what I say is it is like how people have now moved from cars with drivers to self-driving cars or driverless cars. You will see the factory also moving to a self-driving factory, basically. Right? I mean, it's not too much distant in the future where you should see your factory completely being automated.

(Erwin at 00:38:07) And planning and scheduling has to be a key for that, for example. Because, you know, you have to be able to react to something and say, "Okay, make a decision now. Oh, okay. Like, you know, this vehicle was supposed to come here at this moment, but no, it has not come in. Okay, what's the next best vehicle should I send?" There is no one out there to make a decision. So you have to have your planning and scheduling software be able to make those decisions for you. And then that decision is sent to the MES system, and that says, "Okay, you know what? We'll send this vehicle here, which is there in your stock or in your buffer," basically. And it keeps your plant running in that sense.

(Erwin at 00:38:53) And that, I think, is the vision of what we at Optessa are having is this—what we call Industry 4.0, right? That's where we see planning and scheduling going as part of Industry 4.0, basically.

(Joel Beasley at 00:39:00) Are you going to miss driving? Like, when in 20 or 30 years, when I think we'll have this time where self-driving, like, is coming out, like, right now, you know, you get some of the full self-driving Teslas, you see them in other countries as well, and then we'll get to the point where, like, the majority of cars are self-driving, then like, all of them. And somewhere in there, there'll be this gradient of like, I can take it off of self-driving mode. And then at some point, there's gonna be like one interstate that like, doesn't let you take it off self-driving mode. Then there's gonna be another interstate that requires the self-driving mode. And then eventually, no more self-driving, and you're gonna end up like at a horse farm, renting a horse on the weekend so you can drive it.

(Erwin at 00:39:43) That's right. That's right. Absolutely right. Right? I mean, yeah, I mean, I can understand in the sense of, say, driving. Right? I mean, a lot of people enjoy driving, and I myself enjoy driving, basically. I still love the manual cars, basically, because it gives you the pleasure of it. Right? So right there, years ago, manuals became automatic transmissions, and now you have very few manuals in the market.

(Erwin at 00:40:10) And so the same thing is gonna happen. You'll have hardly any people driving. That's gonna be good and bad. Right? The bad, of course, is people who enjoy driving will not have the pleasure of driving in that sense. I hope maybe they can, you know, they'll go to video games maybe and learn to and do driving with video games maybe. I'm not sure. But, like, what may happen with that is maybe accidents may reduce, right, in that sense. Because, typically, accidents happen because there's always someone who does not follow the rules. Right?

(Erwin at 00:40:42) So in this case, like, maybe because they are self-driving vehicles, assuming everything is going on well with the sensors and everything else, there is no issues there, you'll have maybe less of accidents. But, you know, but then, definitely, the pleasure of driving will probably reduce and go away in that sense.

(Joel Beasley at 00:41:02) You know what's interesting? Like, as you're talking, I'm thinking about this. The thing that one generation has to do—right? Skip a couple generations, and it's what that generation, like, almost longs for.

(Joel Beasley at 00:41:16) Exactly. Like, you used to have to ride the horses, and everybody would hate it because it was so gross and dirty. And like, everybody—and then they got the cars. Everybody was like, "Woo-hoo! We get to drive now." And then everyone started being nostalgic, and then only the, like, wealthy people end up having horses on their farms.

(Joel Beasley at 00:41:35) And I heard somebody say the other day that one day, being able to, like, come down to Earth and, like, go fishing or come down to Earth and, like, go to the beach or something, that's going to be what only the wealthy people can do because everyone's gonna be, like, in space. And to come down to Earth will be, like, a big deal. So now on the weekends, when I'm spending time with my kids and, like, I'm going outside and going to the park and doing things, I just am thinking, hey, I'm a person in the future. I'm living 200 years in the future, and I'm, like, one of the wealthiest people in the universe.

(Erwin at 00:42:08) Right. Exactly. Well, I mean, we can see that, right, with your with the kids also. Now especially nowadays, right, with the with the COVID situation and stuff, lot of schooling is online and stuff. So the kids have a lot of screen time. The whole time, they're on the screen. I mean, even before that, people were all the time—at least the kids were, you know, I mean, we tried to teach our kids not to be so much with the screen time. But, I mean, you would see a lot of people—like, because of social media and stuff and things, so they're constantly, you know, glued to their phones or the tablets and stuff like that. Right?

(Erwin at 00:42:41) And people are missing out the joys of going out to play, interacting with other human beings, basically. Right? I've seen examples where even within the family, like, you know, the family says, "Okay, dinner is ready" by sending out a text message or a WhatsApp message to their family—"Come for dinner" or something. Right? It's sad to see in that sense, like, you know, that we are losing the human element of some of these things. So I think it's a—we need to make sure that we have a good balance of these things, right, in that sense.

(Joel Beasley at 00:43:13) Have you seen the TV series Year Million?

(Erwin at 00:43:19) No, to be honest, no. Because I—

(Joel Beasley at 00:43:22) Okay. Right. So you're gonna—I don't watch much TV, but I like the documentary type stuff, though, when I do. And it's a National Geographic series, and it's called Year Million. And it imagines the future in large chunks of time from where we are now to a million years in the future.

(Erwin at 00:43:40) Right.

(Joel Beasley at 00:43:41) And it deals with some of these things, like this family, and they go back and forth between, like, experts talking about it, and then they dramatize a little bit, kinda like The Social Dilemma if you saw that documentary. But they, one of the things that they were dealing with was all of the family was, like, neurolinked up except for the dad, and he didn't wanna, like, do it. And he was upset at dinner that everyone's, like, neurolinked talking to each other and not talking out loud, and then he finally gives in and gets it or whatever. But, yeah, there's a bunch of different, like, I think, like, 10 different hour-long episodes, and it's absolutely fascinating just the way—just to, like, stretch your mind sometimes and think about all the different things that are happening at the same time with regards to our technology just massively expanding.

(Erwin at 00:44:29) Exactly. That's true. Right? I mean, it's a fine balance, really. Right? I mean, and that's what I try to teach my kids too. Like, I mean, make sure that you have this balance. Right? I mean, otherwise, you may just, like, I mean, go completely, like, you know, get lost completely in one thing, and then you will not be able to, like, ever enjoy what it was meant to be a kid, like, you know, to go out and play, to have fun like that, like, you know, in that sense. So, yeah, it's important.

(Joel Beasley at 00:44:59) They're gonna be telling their kids, though, "You have no idea what it's—you need to be swiping because you're using just the Neuralink. You're so—you don't understand the joy of swiping."

(Erwin at 00:45:08) There you go. Yeah.

(Joel Beasley at 00:45:09) And the play button. You're just using your Neuralink, right?

(Erwin at 00:45:13) With my with my kids sometimes, like, even when it comes to, you know, writing a WhatsApp message, sometimes they use their help because they know what it means to right-swipe or left-swipe or something and all those various things. I don't know all these things, and I feel a little, "Oh, okay." Like, you know, so they know a lot more than what we do when it comes to these type of things. And you can imagine what you're asking.

(Joel Beasley at 00:45:32) I have nephews, and they were over at the house, and they brought their, you know, video game console. And they were playing some game, and I was like, "Oh, let me jump in and let me see this," because, I mean, I spent hundreds of hours a year gaming, like, you know, back in the day. And that wasn't even, like, 10 years ago. But, like, they're so complicated. It's like, how do you know what's going on?

(Erwin at 00:45:58) Exactly.

(Joel Beasley at 00:45:59) There's there's no way this whole generation can both have ADD and understand what's going on in these video games at the same time because you need, like, next-level concentration to understand the—

(Erwin at 00:46:12) Exactly. Right. Yeah. That's true. Yeah.

(Joel Beasley at 00:46:12) Oh, man. This is great. I really enjoy talking to you. Have you been—do you follow, like, OpenAI and what they're doing there? Do you get excited about that type of stuff?

(Erwin at 00:46:20) Yeah. So, for example, right now, like, just to say, like, you know, in terms of what Optessa is really, like, you know, looking at. I mean, one is, of course, I just sometime back talked about AI and, you know, how important it is for us. But even, like, you know, for example, you can talk about in the operations research field, one of the new things is semidefinite programming, which is, like, you know—that is something what we are also trying to follow. Another important thing for us, by the way, is—

(Joel Beasley at 00:46:47) What what is that? I'm sorry to interrupt you, but I've never heard that before.

(Erwin at 00:46:50) Oh, okay. Semidefinite programming?

(Joel Beasley at 00:46:53) I don't know.

(Erwin at 00:46:54) Okay. Yeah. It is because, like, you may have—we have heard of linear programming and integer programming, nonlinear programming. Right? So that's the next big thing for solving convex problems, basically. Right? So, things like that—we have to look. And as I said, right, we have to always be aware of what these new technologies are coming up because that helps us stay ahead in the game. So we also have to be aware of what is happening on the computing side. For example. Right?

(Erwin at 00:47:21) I mean, when I first started off, by the way, the machine I was working on was an x86 processor with two MB RAM. How many people know about those things these days? And when I came to Canada, the machine was a Pentium 16 MB RAM. And then I was so happy. "Oh, wow. Like, you know, I could actually solve a problem now," you know, in the sense because it had computing power. And today, one of the reasons we are able to solve this problem is because of the computing power. Right? So that's another thing we're keeping a watch on.

(Erwin at 00:47:54) Nowadays, computing — a few years ago, what we did was when CPUs started having these multiple cores and stuff, we were able to enhance our underlying architecture to be a parallel computing type of architecture. Right? We could have multiple threads and parallel processing and stuff. And because of which, we were able to reduce the time taken to come to a solution by three times, basically.

(Erwin at 00:48:21) There was a huge gain in the performance. So now what we are looking at is we see a lot, especially in the AI ML field, we see a lot of GPU-based computing happening with CUDA and things like that. Right? So that also is something we have to keep an eye on because that can be another game changer, and it will allow us to solve larger and larger problems. Now I'll tell you what you mean by larger problems.

(Erwin at 00:48:45) Right? I mean, typically, when someone makes a sequence, you'll say, okay, I'm making a sequence for a day or I'm making a sequence for a week. Just in today's call, I was having this call with this customer. They're talking of month-long sequences and stuff.

(Erwin at 00:49:01) And we're asking month-long sequence. Right? I mean, as you're trying to sequence 30, 40,000 vehicles sometimes, why do you want such long sequences? And then that goes back to say, okay, the more I can give something to my supplier in advance, the more is the stability I have with the supplier.

(Erwin at 00:49:21) Because sometimes, one of the main reasons OEMs suffer is because the suppliers are not able to supply the parts on time. So the more advanced notice they can give them, we can do a much better job. And Optessa does a very good job because we have something which not many people have, what we call resequencing with bias, basically. Right? What we say is that once you sequence something, the next time you sequence, we will keep a bias to your previous sequence so that your change points are much lesser compared to your previous sequence.

(Erwin at 00:49:50) So that means the suppliers do not see a huge ripple effect of ups and downs. Initially, we told them, oh, tomorrow you have to supply 50. Now we're saying supply 500. That won't do. They will not be able to handle that.

(Erwin at 00:50:04) Right? So now, as people are trying to extend the problems they're trying to solve, that becomes a challenge for how we can get those things to solve in quick enough time. And gone are the days where people were ready to wait for a whole day for a sequence. They say, no, I want it done within an hour.

(Erwin at 00:50:24) Because a person right now, especially a production planning and planner and scheduler, is not going to do just one thing. Right? In the olden days, you would have in the department maybe three schedulers — three schedulers per plant. Now you have one scheduler doing three plants, basically, in that sense.

(Erwin at 00:50:44) So obviously, you're looking for something very fast, very quick. So these things become so very important for us that we have to always keep an eye on these things. And not only that, finally, we are an IT system. Right? It's also the manner in how we deliver this solution, basically.

(Erwin at 00:51:01) As a number of orders, a number of features grow, it becomes a huge impact on the architecture of the system. How can we handle that? How can we — you know, how can our underlying database or the underlying software or the code or whatever we do, how do we come up with those improvements so that people can still handle these large amounts of data, basically, in quick time? That's always a challenge.

(Joel Beasley at 00:51:28) So how did you — earlier you said you would tell me how you ended up in Canada, and I was watching the time, and I want to make sure I heard that story. How did you end up in Canada?

(Erwin at 00:51:35) Oh, yeah. Okay. So I was saying that after my master's in IIT Bombay, which is — actually there, one of the problems I did for my master's, I solved as part of my dissertation, was for an ad agency. Right? They have various customers, and they have a budget, advertising budget.

(Erwin at 00:51:59) And the idea is to maximize their TRP. TRP is your television rating points or the target rating points, basically, that you have. And so this ad agency had a lot of big customers and a lot of big budgets. And they wanted to make sure that their ads went on various programs. Right?

(Erwin at 00:52:15) So there are various requirements there. So whether it is, you know, I need to maximize the goal of the reach. Right? I need to reach the whole country. Now India, you know, has more than 20 languages.

(Erwin at 00:52:28) Right? So you have more than 20 different types of regional programs. So if you want an ad to reach the whole of India, you need to make sure that your ad goes out on these various different programs, for example. Right? And at the same time, there are a lot of other constraints which come in.

(Erwin at 00:52:43) Right? So one of the constraints, for example, would be you don't want to advertise any women's products on, say, children's programs. Right? So any sanitary products or hygiene products on children's. Right?

(Erwin at 00:52:55) Because you're not going to get any benefit out of that, basically. So things like that. So this ad agency had a problem. And at that time, I was doing my dissertation in things like underlying genetic algorithms and stuff. And so we used this thing.

(Erwin at 00:53:13) Right? And these algorithms — and they had their own in-house heuristic, which they already had — and we showed them a huge improvement and a benefit. And when we went to them and said, hey, you know, this is what you could do. They said, yeah, cool, very good. And I was expecting that they would give me a job offer then, right, basically. And they actually basically said this, you know, well, you know, I would have loved to give you a job offer, but, you know, the problem is this is too complex for me, what you have done. And tomorrow, if you leave, how am I going to replace you?

(Erwin at 00:53:46) So, you know, I'll stick with what I have. And so that was pretty disappointing, basically. But at the same time, you know, I had spoken to others in the industry, and, you know, they kept on saying, you know, you guys are solving these problems. I was telling them what we do in our course. Right?

(Erwin at 00:54:01) And they'll say, well, you're doing these problems. You're talking of optimality and stuff. You know, forget about optimality. We are running at 70% utilization. Can you give me 10% improvement, basically? And so that's actually — that led me, you know, saying that, okay, let me get in the industry. Let me see what I can do. And so I joined that one company, as I said, that Tata Unisys at that time, because they had this group called an ATG group, or Applied Technology Group.

(Erwin at 00:54:29) And one of their divisions was data mining, expert systems, resource allocation, and scheduling. So I was there with them for around three years. I loved my time out there. No doubt. I managed to get a lot of exposure to some real-world problems, whether it's train scheduling or crew scheduling or even trying to schedule ships coming into the port to minimize damage, basically.

(Erwin at 00:54:51) Right? So, you know, you get a chance to look at all these things, and it broadens your scope. You get to know so many different things, which otherwise you wouldn't know sitting in an office, for example. One of the cool things was, I mean, when we did the railway problem, we went and visited these various stations and the various station masters. And one of the station masters said, you know, you should go and talk to the person who is manning the railway crossing gate.

(Erwin at 00:55:18) And we were saying, a person manning the railway crossing gate. Why would that be? You would think it's automated. Right? At the moment, a train is coming, it would sense it, automatically close the gate. But when you go and speak to the person, you understand the reasons. For example, he said, you know, some places in rural India, there would be a farmer who's going with the bullock cart, which is heavily loaded. And he may be going so slow or he may get stuck on the tracks. If I have an automated gate, that guy is going to get run over, basically.

(Erwin at 00:55:46) So we cannot afford to have these type of things in certain places. Some places we do. So, you know, you get these types of insights into the actual problem. And, you know, you need to try giving solutions to this type of problems, understanding what the user really wants and not just trying to push a solution. But anyway, so that basically was my motivation.

(Erwin at 00:56:08) Right? Just how can I, you know, give solutions to customers which really means something that are usable for them, not just for the sake of having a system? But after three years in Tata Unisys, I felt that, you know, the sales cycles are too long. Being a services company, there's a problem. I needed to get to a product company.

(Erwin at 00:56:28) And so, unfortunately, in India at that time, we didn't have any of those product companies then. So yeah, I started searching for jobs. And one of those — it was a time, you know, when, if you remember, APS and FCS were those keywords were being thrown around that time. i2 was a very big name at that time, i2. And one of the competitors of i2 was ShivaSoft. One of the founders of ShivaSoft was Vasu Srinivasan, who was the founder of Optessa, actually.

(Erwin at 00:57:00) So I applied for the job. ShivaSoft, they were looking for an algorithm developer. Right? And that's how I got this job, basically, with ShivaSoft. And the thing is, I got the job, and I had to be there on 01/01/1999 was my start date.

(Erwin at 00:57:16) Now you can imagine. Right? I mean, I'm — it's like I said the temperature here is plus 30, and I get this job. I don't know where exactly I'm going, to be honest, because I'm going to Edmonton. Typically, when you're in India and stuff, you know, when you say Canada, people only know Montreal and will know Toronto.

(Erwin at 00:57:33) No one would know where Edmonton, Alberta is. Right? And so I get there, and the plane is coming into land into Edmonton, and all that I see is white. I see no houses. I see nothing.

(Erwin at 00:57:46) And I'm thinking, where am I coming? Right? I mean, I see only white stuff here. Am I coming to the right place, basically? And when I land, the temperature is minus 25.

(Erwin at 00:57:56) Right? And I was — oh, and the jacket. I'm not prepared for that weather, basically. So it was fun. Right?

(Erwin at 00:58:03) I mean, and it's just the pure excitement of coming to a new place that kept me going. But really, and atop it all, one of my colleagues from the office on thirty-first night takes me out and says, you know what? Let's go and watch the fireworks. And then at midnight, we're going out in minus 30 degrees Celsius to go and watch fireworks. And I was like, what?

(Erwin at 00:58:22) Is this your sense of fun? Come on. So that's how I ended up in Canada. That's how I started working with Vasu. And then when Vasu started his own company after ShivaSoft merged with another company, I started off Optessa along with current CEO, Ashok. That's when I also then joined Vasu and Ashok with Optessa. And then from then, since then, I've been with them, basically.

(Joel Beasley at 00:58:47) Oh, it's such a good story. That's good. Through relationships, building relationships, and growing together throughout your career. That's a great story, my friend.

(Erwin at 00:58:55) Definitely. And I really enjoyed working with Vasu, working with Ashok, and everyone else, basically. Right? And that's the beauty of it. That's what keeps us coming to work every day, basically.

(Erwin at 00:59:06) Right? It's just this relationship which you build. Unfortunately, when we had the — in 2007, right, we had this whole economic downturn, right, because of the Wall Street, whatever happened. Right? I mean, that's when manufacturing got hit pretty badly. That's the time when also Optessa also had to suffer because of that time. Right? I mean, and some of those — it was very, very sad for us to see some of our people go at that time because there was no other option. Right? I mean, and they themselves realized there was no option, and they themselves decided to find somewhere else or whatever it is.

(Erwin at 00:59:41) A lot of them, you know. But, you know, those were the relationships which we really built with these people, which actually are things which you keep in your heart every day. Right? I mean, you come to work, you do your work, but, you know, it's the people relationships. And that's what I sometimes say — I tell my kids.

(Erwin at 00:59:57) Right? I mean, yes, you have so much on the screen. You're doing so many things on social media. But it's just people relationships are very important. Right?

(Erwin at 01:00:04) And make sure to never, never, never lose them.

(Joel Beasley at 01:00:09) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.