 Welcome everyone and thank you for joining. This is Laura Llege, one of the course leads here at the MIT CDL for the MicroMasters in Supply Chain Management program. I'm very happy to be co-hosting this live event with Mr. Kellen Betz, hi Kellen, also a course lead here at the program. We're very fortunate to have Dr. Christopher Mejia joining us today. This Mejia holds many hats here at the center. He served as the director of the MIT Scale Network for Latin America and at the Caribbean. He's the director of the MIT Graduate Certificate in Logistics and Supply Chain Management, the one we call GSLOG, and that is one of the pathway for MicroMasters in SEM program, and he's joined us today with his research hat, Asif Pounder and director of the Food and Retail Operations Lab. Welcome Chris and thank you for joining here and our audience today. Well, thank you very much for having me guys. So it's a real pleasure for me to be with you today. Okay. So should I get a- So I'll launch a brief poll and we go through the agenda for this session. So if Lisa here from Combs can help us, thank you Lisa for being here today. We want to know why you're here today joining us. So the first could be that you're actually joined for network design because we know many here are on that course right now. Are you here to learn more about AI applications in the supply chain, social sustainability, generally expanding the supply chain knowledge, or just getting a lot of analytical skills and willing to know where to apply them. So thank you for that. While we populate this poll, Kelly will share the agenda for the session. Well, thank you Lara and welcome Chris. Pleasure to have you here today. So for the next 20 minutes or so, Dr. Mahia Chris will be discussing recent work in the Food and Retail Operations Lab he's been working on. We're very excited to bring him in and hear some about recent research that they've been working on, specifically about increasing efficiency in food supply chains to improve access to quality nutrition, especially in underserved communities around the world. So we're very excited to have this discussion today and Lara and I will have some kind of questions be prepared for after the presentation, we'll focus on those for a few minutes, but we definitely will save some time for your questions there in the audience you out there. So please start thinking of those as we progress along in our presentation or Chris's presentation, and please use that Q&A feature there in the Zoom that Q&A button there along the bottom to ask your questions. We'll definitely keep track of the Q&A. It's hard for us to keep track of questions in the chat because there's lots of other comments in the chat. So please use that Q&A feature and be sure to be logged in. We won't pick anonymous questions. So with that, maybe let's check on our results from our first poll here. If we could end that poll and then see the results here. So the question was, why are you here today? I think there's lots of learners, obviously that's awesome. Let's see learners here looking to learn about network design is reward applications. That's great. It seems to be one of the more popular ones. Also, I want to learn about AI applications in supply chain. So that's also awesome as well. And then we have lots of SC4X learners out there where we're studying machine learning and AI. So hopefully we'll touch on some AI concepts. And Chris, I don't know if you have any thoughts about the poll results there. Well, this is quite varied, right? So I will try to address most of them, but well, probably a few of them are not going to be necessarily with too much debt. But I'm sure that with your help and the other life events that you will have, you can expand a little bit further. Awesome. Well, with that in mind, do you want to kick things off and take over here, Chris, and start your presentation? Sounds great. Sounds great. Well, as I was telling you, thank you for having me, guys. So let me start sharing like the presentation mode. So we should be working probably right now. So, well, hey, everyone. I really hope that you are going to enjoy this. So hopefully you can provide also feedback at the end of all of this part. So let me start by telling you some interesting figures. Together started the world in the previous years as a food system produces or has produced around 11 billion tons of food per year. Imagine that. And this space was maintained during the pandemic, but the supply chains have been super resilient. And something that has helped a lot is the supply chain network design behind all of them. And today I'm going to be discussing another type of supply chain network design that is social driven because we want to bring nutrition to the people who need the most by using similar techniques like the ones that you have learned or you are learning at the SC2X or at the SC4X, okay? So as you will see, most of my dissertation today is focused on social driven supply chain network design bringing nutrition through all of these AI components to underserved communities. And this is a work that I typically do with many colleagues. In this case, I had the opportunity to collaborate with an SCM master student, the PhD candidate. Nowadays, Sanchita Das, who is studying her PhD and the University of Washington on the other coast and also with my colleague Tatiana Sadala, who is a nutritionist by training. Well, anyways, let's start this journey, okay? So together started, what is the food and retail operations lab or what does it exist? So in short, so we can speak about many different food issues around the world, food malnutrition, food insecurity, food waste, et cetera. But if you think about it, this is a multidisciplinary problem and that is addressed by supply chain managers like us that we are studying this by economists and also social scientists because they really need to understand what is going on with the price and everything else, right? So if the price is not proper, food is not affordable and the rest of the supply chain is gonna fail. Innovation science is because you need to make sure to create some kind of a wariness, kind of a tool, a technology that can help, like bringing more people to these vulnerable population segments. And health science is because it's everything about nutrition, right? So you are not taking proper food. This is not gonna work properly. But the food and retail operations lab works in these five dimensions that you see at the center. So first we want to address everything related to combat food malnutrition of any kind. Food malnutrition is a wide word but basically involves or comprises food insecurity, obesity and overweight. So naturally this is related to nutrients and you need to make sure that we bring proper food to all the population segments. Second, we want to reduce food waste and food losses through circular economy approaches and many other innovative schemes. Then we need to make sure that this food is edible and has the quality that we need in order to eat it. That's called food safety, okay? That is different from security because food safety is more related to reducing any kind of breaches because there is a cross-contamination, for instance. And the previous two last component is how you connect the small holder farmers to the rest of the ecosystem. And this is part of what we're gonna be talking today. How do you connect the small holder farmers locally to combat food malnutrition, okay? And the long term, let's say a next frontier that we have for the lab is how we can transform this into long-term sustainable food ecosystems through logistics, okay? With all the different strategies, operations, understanding human behavior, the technology as a driver and to provide proper governance, okay? So this is a few of the countries that we are working at with different projects but that's something that I'm not gonna elaborate too much today. But first, I want to really make you think about the following. And by the way, I'm gonna be using something that probably you will learn in SC3x that is actually a system dynamics, okay? So you haven't taken that course, well, bear with me, okay? I'm gonna make it like easy for you to understand, right? But why are food supply chains so important? So let's take one specific issue in the world, food malnutrition, okay? I already mentioned food insecurity, overweight or any kind of formal like that, okay? So if we try to relate food malnutrition to economic growth, so we can create some kind of a polarity. So as you can imagine, higher the food malnutrition, what do you think is gonna happen to the economic growth? It's gonna be lower, right? And then if we relate this to productivity at work, higher the economic growth, higher the productivity at work. If we want to relate this as a person impact in the rest of the ecosystem, we need to understand the country's performance, the country's economic performance. So higher the productivity at work, higher the country's performance in the economics, right? And then we can continue adding poverty rates. So if you have, and probably it's better if I use like the last pointer here for you guys, higher the country's economic performance, lower the poverty rates, right? But higher the poverty rates, oops, sorry, higher the food malnutrition. So if you remember from your elementary school or your arithmetic classes, you can multiply the polarities and then you will get a sign that is positive, right? Minus, plus, multiply, times plus, times plus, times minus, times plus is positive sign, right? So this is a reinforcing loop. This means that it will grow without control. And this means that the malnutrition will put the brakes on the economic growth. And we can do exactly the same to understand any other type of problem here. You can call this like poverty. You can call this like food waste, you know? And this is creating an issue at the economic level. But guess what? The same happens to social part, the social development. It's also an issue and we can create similar polarities and then we will find another reinforcing loop that is called social backwardness. And we can do the same if we think about our own health, right? So, and if we speak about our own health is another reinforcing loop and this is the health burden. So in other words, food malnutrition is gonna be impacting the economic growth of a country, the social development of the country and the health of the whole country. And as we know, education is in between the economic growth, the social development, the health and the productivity. Therefore, this is very impactful. We need to make sure that we are eating proper. That's one of the reasons why. But now let's start talking about other things because unfortunately this is creating wrong dynamics. This kind of a vicious circle that we experienced here because the more that we start eating wrongly or other types of goods that are not necessarily providing nutrition, the more that we are getting rid of these fruits and vegetables perishable products that are bringing nutrition to the neighborhoods. For instance, in this case, I'm gonna read this with the opposite sign, okay? But bear with me. So lower the fruit and vegetable affordability. So this means that you don't have enough money to buy these fruits and vegetables. There is gonna be lower demand, right? Lower the demand for these perishable products, lower the accessibility. Nobody will like to put them in the shelves or make them available in their corresponding stores, right? Higher, sorry, lower the accessibility for fruits and vegetables, lower the demand and lower the demand, lower the, well, the affordability is gonna be also impacted, right? Because if somebody wants to sell them, they are gonna sell them at a very astronomical price. Therefore, this is not good. And this is typically what we call not only in developing countries, but also in the United States, food deserts, okay? This also happens in the United States. And for the record here in the United States, with $5, we can buy 312 calories like this. So it's a bunch of grapes, two broccolis and some orange juice there, right? But for the same amount of money, we can buy this quantity of calories. Are they nutritious or not? Well, that's something that's gonna leave on you, leave on you, but naturally we can improve it, right? We need to make sure that we are, you know, like getting proper nutrition to our body. But now let's start thinking about what is the undesired effect of all of this? For all the food that we are not necessarily consuming, all these tomatoes that we have at the fridge that we buy and we never eat because of several reasons because we don't have the time to cook it, et cetera, we are generating plenty of waste. In the United States, one of every two apples goes to waste. In Latin America and the Caribbean, one of every three apples goes to waste. And in the case of India, it's four out of every 10 apples goes to waste. And with the same quantity of these food waste that can be recovered, we can feed this quantity of people, 15 million inhabitants living in very poor conditions in the case of the United States, 36 million in the case of Latin America and 12 million in the case of India, right? And remember that the third largest polluter in the world is actually food waste. So if we start recovering more food and making a difference, this whole ecosystem is gonna work better and benefit the environment too. I'm not gonna speak about like the last column here for the sake of time. But now let's start discussing about like the topic that we want to discuss today. So first, how many of you know what is this? This is a superfood and it's called chia, okay? And there is another one. This one is quinoa, okay? And the last one is called tarwi or Andean chickpea as they will call it in the United States and in English speaking countries. Well, all of these ones are superfoods, but how many of you guys know that some of these foods are not necessarily, let's say palatable or accepted in some of the cultures? Some of the cultures wouldn't like to eat chia, quinoa, or actually tarwi, because the religion or many other cultural issues are gonna be affecting. So all of these matters. And as you have learned probably or hopefully with all the micro masters, right now you know that who is at the end of the supply chain is the consumer. And if the consumer changes everything, everything is gonna be affected. So we need to make sure that everything, supply chain network design and everything else is gonna be aligned with the rest that we are preparing, handling, distributing, okay? Well, what is the case of India? We know that for in this case religious reasons, most of the times the beef is not eaten there, okay? Because the religion, the same happens with some of the pigs, the porks, right? So the vast majority of the plate that they are eating every day is based on these 50% fruits and vegetables, 10% is pulses, eggs and other flesh foods, 4% is fats, oils, nuts, milk and 23% is cereals and other nutritional cereals is like a millet, okay? So as you can see a vast majority of what they should eat is based on plant-based protein, right? But something that we discover because at the end we, is that there is an eternal divorce between nutrition and operations research. If you remember what you learned in SC0X or maybe you will learn it, it's a very good example of how to learn optimization and it's called the diet problem is where you need to minimize the cost and then you need to satisfy certain restrictions larger or equal to something. Do you remember that? Well, let me tell you the following. In reality, our body doesn't behave like that. It's not like, oh, I need to eat three apples or because this is gonna bring me like, I don't know, 300 calories and then I'm gonna be able to absorb these vitamins is not like that because once we cook the food or if we are sick the way that our organism, our metabolism is gonna be processing all of this is completely different, right? So we need to change the approach and this is exactly what we did here. So what we discovered while checking something that exists in India that is called the public distribution system that is a system that was developed in the early 60s in 1960. They started like bringing mainly rice and wet to the population segments in order to start reducing the food malnutrition. But something that we discovered together with Sanchita is that it's a great effort but in terms of cereals, they are bringing more 51 kilos for a family of four when you actually need 32 kilos. And in the case of the pulses that is the plant-based protein they are only bringing 3.67 kilos and then they need actually 10.8. So it's only 33%. So there is deficiency. So there is malnutrition, right? So all the project is about how we can bring more plant-based protein to all the population segments especially the ones who need the most, right? And the research question is because remember everything that we do at CTL is research oriented. So question number one how we can design a healthy, affordable, locally sourced food combinations or food baskets that Indian households can buy? Second, how we distribute all of this? What type of supply chain network design we can model in order to make that happen? So as part of the answers what we want to solve is provide more available nutritious food at affordable prices especially for the ones who need the most and provide this accessibility to all of them just to make sure that all the geographies throws India, they have access to this at a proper price and these are available to them. And well within the scope what we are modeling here is cereals and pulses. We haven't modeled yet fruits and vegetables that's something that is coming. We are using only the public distribution system that is managed by the government. We are not changing that. And our main target is like the poorest segments of the population, okay? We are serving all of them but we want to focus particularly on those. So the methodology first we wanted to segment all the customers because all of them are different. We consider the geography where they are living in what region, West, East, North, South, Center, okay? So economic features, how much money they are earning? What is the average income level? What is the population density? What is like the historical consumption of different types of products? And this is the demand and the taste preference that is coming from an expenditure data, consumption expenditure data from the government. So based on that, we started using PCA and K-Means clustering just to understand how these districts look like, okay? And how these clusters look like. Then in the second part, once we characterize the customers, we want to characterize the food baskets. In this case, like the grain and cereal baskets. And what we want to make sure is that the assortment of each of these baskets is actually affordable, well assorted and nutritious. Also bringing a plant-based protein, you know? And for that, we use a beanpacking algorithms to make sure that this is possible. It's similar to the knapsack problem. Do you remember that from the SC0X as well? So you want to put as many, you know, not nutrients in that knapsack, right? In this case, it's a bean, right? With other characteristics to make sure that you are bringing nutrition to that specific community or that specific district. And last but not least, well, what we wanted to hear. So what we want to make sure is that we are building a mixed-integral linear programming that is multi-commodity and is handling these different types of cereals or baskets across the distribution network, okay? We want to make sure that we are counting for those storage locations that are like silos from the government and we are bringing all of these to the hands of the really needs it. But in order to understand this a little bit better, it's probably easier if I show you what is like the steps that we follow. So a traditional farmer in India and in many other countries, right? They can decide if they want to sell these products directly to the end consumers or they can sell them directly to the government. In this case, what we're gonna be paying attention to and what is within the scope is all what you see in this square, okay? So the farmer sells this to the government and the government is gonna make sure to bring this to a collection center, a procurement center. And then from the procurement center, the government is gonna ship this to a storage center. It's kind of a, it's a coupling point in this case where they are gonna be bringing different types of products, particularly a repeat rice and wet, okay? And they are gonna be transporting this by road. And then once they go to this big silo, they are gonna be distributing this across India, among different states by rail. And that's why the rail is very important in India in order to bring all of this nutrition to the ones who need the most. Because as you can imagine, well, as in any other country, the weather conditions are different in different states, in certain states are more arid than others, other ones are a little bit more humid. So we need to transport like this food and life to the population segments. And then this is receiving the other silo in one of the states and then the distribution process starts. So from the silo, they are gonna bring this by road to the distribution center. And from the distribution center, the government is gonna develop these fair price shops where they are gonna be actually delivering this to the end consumers, okay? The model looks as follow, okay? So I'm not gonna deep dive for the sake of time on all of this, but I want to give you an idea that what we want to do is to minimize the cost for the government, including the procurement cost because naturally this is not for free, yeah? Second, we need to minimize the storage cost and the handling cost. Every time that we are passing this through different hands, from the procurement center to the storage center, then to the storage center to the storage center too or the other silo, and then to the distribution center that is the decoupling point, we need to consider all of that. And of course, all the transportation, right? In between the different facilities. Key parameters, the consumption that each of these families, households do, right? Or have, how many households do we have per district? In India, the government has divided the whole country in 625 districts, okay? For which they have a lot of very granular information related to the customers. And this information is publicly available, okay? And then what we want to really understand is how remember in supply chain network design you need to pay attention to the demand. We are talking about the demand, you need to pay attention second point to the capacity and of course to the locations, right? And that's exactly what we are doing here. And then we are paying attention to how much of these produce items can be grown locally or are sold directly to the government. But it's not only wet and rice anymore. Now we are adding plenty of millets, pulses, et cetera that I'm gonna show in a moment. And of course, related to the constraints, well, the first two takes preference and demand constraint. Well, we need to warranty that we are meeting the demand, supply constraint and flow balance constraint. We need to make sure that we are like acquiring, procuring the product and bringing it to the hands of the end consumer up to here. And then we need to make sure that we are respecting the capacity of each of the different facilities, okay? That's pretty much it. And I didn't write down here like the domain constraints, but remember that they need to be, you know, like also larger or equal than zero, okay? So now the results. And this is where we started using plenty of AI, right? Especially machine learning. So, clustering analysis. First we define like how many clusters we really need it, right, especially for the food baskets. And then we started, you know, like using this elbow method just to make sure that we were like with a value lower or equal than the agent value equal to one. And this is a value of six typically. So then we started building the clusters for the consumers. Do you remember that? Based on the region, based on their socioeconomic features, et cetera. And then we started paying particular attention to what they were consuming in each of the different regions, you know, or by each of the different clusters. And something that you immediately realize is that there are some peaks or maximums that are rich, for instance, this cluster is eating a lot of small millets, right? And the biggest cereal that they are consuming is actually rice, okay? Well, this cluster is completely different. It's a little bit more steady, right? In the maximum that they are consuming and the peaks are at these different lentils that they have, mug, oar and besan. All of these lentils are like pulses, right? Or beans that they have there. And those are different types of plant-based protein. Okay? So once we characterize like the people that we have here and also like the clusters of food that we have, we need them to understand like how many clusters of food baskets we need, right? So we define six of them. And this is like the big summary. As you can see, like the vast majority of the content is rice and wet. We want to preserve that. It has work is bringing nutrition to the inhabitants there and the population, but we also want to make sure that we are bringing something else. And this something else is plant-based protein that is locally grown, okay? And then this is exactly what we started bringing. So in the case of cluster number six, for instance, we have around 18% of rice where it's around like 52% or less or 50%. Then we are bringing our heart that is these poles that we have here, the one in yellow. Then we have a jaw work and then we have bashera, okay? And these are the ones that are bringing like a little bit more nutrition. And don't worry, guys, it's just two more slides and we are gonna be ready for questions. So start writing your questions there, guys, okay? And then, so probably the main question is, okay, perfect. So you already understood what is happening with the consumption and the demand you already understood like how you can build these different baskets, right? Based on the location and many other characteristics and the nutritious content that they are bringing. But now how you are gonna distribute this? That's where the supply chain network design the mathematical model, the optimization model that I showed you is gonna be playing an important role, okay? So this is the picture before our model. And well, for the ones who are not from India, this estate that you see in red is the one that is, let's say, accumulating most or concentrating a lot of production and also distribution is called Punjab, okay? So you like Indian food, there is very good Punjabi, you know, like food, right? So, and this estate is actually followed by others like this one, this one is called Uttar Pradesh. She's one of the most, is the densest estate in terms of population. We have others like this one that he, oh, I forgot this one, sorry. This is called Haryana. That's where the national capital region is located where New Delhi is located. And this other estate is called Hiderabad, if I don't remember well. But as you can see, we're bringing all of these but the procurement and the distribution happens at a lower scale, right? Especially in estates like Gujarat, that is this one here that is completely white and this one that is called Maharashtra. So, and Maharashtra is a state where the capital is Mumbai, okay? And this small estate here that is called Gua. So in other words, we are not necessarily bringing like the quantities that these estates need. So when we run our model, including not only rice and wet, but also other protein-based pulses, look at this. So we were able to balance this a little bit more, okay? And I know that several of you would say like, okay, and that's great. Now we can see that Punjab is a little bit more balanced. So you were able to increase like what you are bringing to Uttar Pradesh, because we went from this light, this light green that you see here to this yellow that is here. So we are bringing a little bit more proportion. And the reason why this is happening is because what I already told you is one of the densest estates. So we need to bring more to this population but the way is one of the poorest areas in terms of wealth as well in India, right? So you are wondering, okay, why this is happening, right? I'm not showing you like ISO weather or ISO terms, like a map, that's where you can see the weather and also like the different climates inside the region. But most of these parties are it. This part has a little bit of a desert here as well. So that's one of the reasons why these estates are not able like to produce their own wet rice and other pulses, you know? So you need to make sure that estates that are a little bit more humid or have a better weather I'm gonna be like bringing all of these products to the others, right? Particularly in the case of rice and wet. The other ones that are locally grown might be enough to suffice, you know, like the suffice or the part related to the plant-based protein. So on just last comment here, we're also bringing food to this estate as well that is called ASAM, if I remember well. And the ones that you see in white are those estates that are under dispute, okay? With other countries. I'm not gonna talk about politics, but our model can start bringing like also food to those is just that by, we put these constraints by requirement of some colleagues, okay? So let me summarize what you just learned today about like the study. So number one, how we answer the first research question how we designed this healthy affordable basket. So we are actually capitalizing on the fact that we are locally growing pulses in different parts of India. And we want to make sure that we create these or we use these artificially intelligence approaches to create these baskets via clustering based on local taste preferences and also how much the agronomy there is helping, you know like to put these products into these baskets as a cost efficient strategy. Second, how we are like bringing all of these to the rest of the population by using the public distribution system that I told you but making a few changes, you know? Part of the things that we want to start doing is to reduce the number of intermediaries and make sure that we are able to move on with other stuff as well, okay? So in order to wrap up here so hopefully you like this idea but I want to finish giving you like four last messages. Number one, improving the food supply chains is very important because as you saw, this is the minimum that you can do to improve the productivity and the economics of account. Second, there is a huge opportunity to start thinking about supply chains not only as a supply chain management based oriented stuff, you need to start like understanding how other disciplines like nutrition, et cetera can start helping create other type of schemes that might be useful to intervene to change that typical approach. So because at the end, this is the third point if you want to really create long-term sustainable food ecosystems, you need to look for holistic solutions. It's not only the supply chain management based one that is going to be the only one that you should follow you need to pay attention to the other ones. And last but not least, we are not, let's say, paying too much attention here to what Kiranas that is mom and pop stores or nano stores as we call them can do, but it's important to start developing this as small to small business environments in order to support the delivery of goods, right? Otherwise, it's going to be insufficient to reach out to the last location that is very isolated in a particular place in India or anywhere in the world, okay? And of course, without this work wouldn't be possible without all my team and particularly the help of Sanchita Das she's Sanchita from West Bengal and my colleague Tatiana Sadala Kolesi who is actually our nutritionist those were the ones who changed my chip and made me realize that, yeah, we're committing a mistake if we are using the diet problem, okay? So that's pretty much it. So I would like to pass the word to you, Laura and Kelly probably to listen a few of the questions from the audience. Awesome. Thanks for sharing that this last slide as well because we have a lot of questions asking where to find this research if you have any papers, publications or any information to share. So everyone here in the audience, if you're interested please contact Chris Mikiya or go to the website from the Food and Retail Operational lab to find out more on the details of the project. And of course you have the video recording in YouTube on the MIT CDL channel. So personally, I felt it was impactful because you showed us the huge impact food malnutrition can have in many different aspects. And sometimes like we've got all these tools and you have seen like, you actually went through all our program, I would say with the different tools you have used and we learned about these tools but probably we didn't think of applying it in a way that you did and how you combine all those elements and knowing that we can make such an impact by using things that we get access to probably in a smaller scale but I think it's also insightful and inspiring for our audience and based on the comments and the emojis I'm seeing popping up in the screen I hope that everyone else felt the same way. So there are many questions but I will start with one about data preparations because we are talking about multimodal logistics including road and rail you're talking about the complexities of politics and regulations. We got the government intervention here. You're talking about local sourcing about the different types of food that are produced and how those help with nutrition. What was the most challenging part of working with all this information combining preparing it and like a little bit on data management process you did? Yeah, this is a great question. Thank you, Laura. I would say that probably the hardest part is not to actually process the data. I think that that's kind of easy but to find the data. And I think that that's something that you will face in your daily lives guys as well. So you need to make sure that you have like some kind of a framework that is useful for you to look at the right ways and the right sources to find this out. So at the beginning with Sanchita Sanchita had this very clear, right? So she told me, I know that this expenditure, this consumption expenditure data exists. Also these records about like how much crops and the yield is, but it's quite fragmented, right? So once we found that, we needed to process and to put this together. So probably that was like the challenging part. But thinking about like the future, especially for the learners and the MicroMaster credential holders, I would say that what I was telling you is true and holds you need to find or create a framework for you to be resilient and say like, what would it happen if I'm not able to find certain data? So my recommendation there would be like, try to categorize or classify the data depending on the urgency and the importance, right? And also how much granularity you need for that data because the preliminary model of this by the way was with hypothetical data, right? And then we started like improving it as we move on and we started looking for, okay, let's look for data related to nutrition. What type of data is available? What we can do? Because originally, if I tell you the truth, we wanted to include like micronutrients, micronutrients, like vitamin D, vitamin A, mineral zinc, blah, blah, blah. So the nutritionist that Tiana told us now is not a good idea. You need to go for micronutrients, right? Because, and that's easier to get in terms of data, right? And that's in terms of data. Now in terms of modeling, because I typically like divide my advice on that this way. So in terms of modeling, I think that the challenging part was to put everything together. As you can see, there are like three different things here, two of them AI base. And despite optimization, now it's kind of starting being considered AI as well. Well, we put it like in another pocket, right? So put these things together was challenging, but we found a way through like a lot of resilience and making sure that whatever we were getting at the beginning was we're gonna be parameters for the optimization model. And last but not least, the lesson learned is like, you need to gain a lot of resilience as you move on with these models. You need to make sure to make some sacrifices because sometimes the perfect model is not gonna be possible. So you need to make sure to be like open to say like, I want to make sure that I get this. This is the objective of my research, but and this is how I plan to help the humankind. So, and this is what we did here. We didn't want to deep dive too much into nutrition components or like technicalities and be more practical. That would be my take, Laura. Awesome. Well, thank you, Chris. I definitely would echo Laura's comments on how it's kind of inspiring to see the application. Some of these tools will be learned along the way, some machine learning and operations research and some of these tools will be learned and maybe kind of a more of an academic context being applied to kind of really important problems like food, not nutrition. I know that's been a problem kind of all probably all of human history, but it's still a problem with conflict around the world, increasing some of that in certain regions and other regions going through drought and deserts desertification and all those kinds of things and climate change in the future probably bringing even more of that in the future. So I'm super inspiring to see that. I want to maybe dive in, kind of build on your comment about the data and dive a little bit more into the customer segmentation. I think it's kind of fascinating to be able to have that level of data for a huge region in the world like that. And so maybe if you could elaborate a little bit further on how that clustering process worked and maybe what were some of the kind of key features of the data that differentiate those different customers. You know, I know you had geography in there and agriculture and all those kinds of things. What were kind of the key features that separated those clusters and the distinct clusters? Yeah, that's a very interesting question. So in short, I could dare to say that in terms of like the main differentiators were probably like the average income level, I would say, and the regions that they were living at. Okay? And if you think about it, it makes sense, right? Because the region where you live at determines because of the weather where you're gonna have available in terms of like the policies and the millions and how much money you have is gonna provide the affordability. So again, it's accessibility and affordability, right? Price and the location of where is this been done. But there was a third factor that for me was quite interesting and it's the guilt. That the guilt of the crops, of course, how much you can grow and it's in terms of capacity. So you are, it's like a news vendor inventory model, right? So if you are over the capacity or under the capacity, so how do you want to handle this? So if you are under and you don't have enough capacity to meet your demand, you need to start looking for other, let's say, states to or like districts to provide you this, right? But in exchange, you need to like provide the polls or the millet that you have. So that's what we found, but this is just to say like what were the determinants or the significant factors. The process was the following. As I mentioned in the methodology, we started with the clustering, right? We really needed to understand like through K-means because all of this data was a quantitative. In case that you want to use qualitative, you will need to use K-modes or you want to use like mixed, you need K-prototypes, right? You know, quantitative and qualitative, we use K-means because all the data was quantitative. And then we determined like that we needed six clusters, six different baskets. And then we started feeding like this information but before we went into the feeding information and analyzing it, we said like, oh, wait, we need to understand what are the drivers? That's why we use the principal component analysis just to determine like what are the factors where we need to pay attention to and what not to? Because if two different, let's say variables were in the same direction, were parallel, you know? Those ones, if you just take one, that would be more than enough. And technically speaking, they are correlated, right? So you just need to pay attention to one to avoid multicollinearity as well in the future to inflate to much certain variable. And then after we did this PCA analysis, that's where we started saying like, okay, these are like the characteristics, these are like the features that the consumers should have. And now we started saying like, okay, this is for the customers or the consumers, the households and the other part should be for the products. And that's where we started paying attention to the yield, to the other characteristics in order to build the food baskets or the grain baskets. And then the process was more or less the same, but in the part of the food baskets or the grain baskets, we needed to use optimization too, because the combination and the quantity of, the combinatorials there are astronomical. You can put like as many pulses as you want, even if it's in a small amount. So we needed to make sure that we were actually receiving a good benefit to include that in the knapsack, right? In the bin, the basket in order to drive it. But there are so many interesting problems that can be improved based on that. Thank you, Chris. And you make it sound a little bit too simple because people here is all willing to start implementing their own models or getting into a startup to help with food and nutrition. So definitely again, into the inspiring side of the supply chain, which I think it's what we are also all passionate about. Just going back into this, thinking on how can I apply something like this? I was wondering if you thought or could try to extend this to a different city around the world, different countries, larger scales or same or smaller ones, different retail formats. And we were wondering if you were exploring that approach if you had anything you wanna share in terms of that. Yeah, I will be brief this time for the sake of time too, right? In order to allow the other question. So in short, yes, we are expanding this project into other areas, whereas you probably know in live, nothing resembles or is exactly the same. So we need to change like many of the things that we were doing here, probably the framework that I explained to you as the methodology is gonna be similar, but the culture, the preferences, the intake, the products are gonna be completely different. So at the moment we're exploring like how to use a similar approach into fruits and vegetables, okay? First in India, we are very interested in star work in actually we're already working with bananas, a little bit with mangoes, things like that, especially for the lassies, if you like them, right? And we are bringing like these learnings also to Latin America and hopefully soon to Africa just to make sure that we use similar concepts. But something that actually intrigues us a lot and I mentioned this at some point in the talk is like these retail formats that you mentioned, Laura, for instance, the Kiranas or the nano stores, the mom and pop stores, they can play an important role and we should not take them out of the other systems. This one in particular, we are using it as it is because the public distribution system managed by the food corporations, that's the institution that managed this in India, it's already there, right? And we are just supplementing or suggesting what they should do in order to improve the delivery, right? But if we think about like how we really want to penetrate to go to the isolated communities where the most vulnerable populations are located, we need more capillarity and the capillarity is gonna be brought by these mom and pop stores that are the small family on retailers that know the community, et cetera, and they know the business, they have a curated assortment, et cetera. So that's what we have been working for already more than 10 years, you know? So, and we are now bringing just the food malnutrition component in order to combat that. And just to wrap up, your first question, you also mentioned like the political part. Yeah, politics, that's another thing that we need to pay attention to because even if the model is perfect and yeah, we need to take into account what is happening in terms of like where the model should be working, otherwise you need to consider other constraints and other things to make it work, okay? Awesome, thank you, Chris. That sounds like a challenging constraint to apply to a model as a political map within a particular area for sure. So, awesome, so we have lots of great questions in the Q&A, thank you for bringing those and keep thinking of those. And we'll start pulling questions from that Q&A maybe before we do while we kind of curate some of those questions. Chris, I don't know if you have any just kind of general comments for our MicroMasters learners or maybe learners who are thinking about the GC log program if you have any kind of recommendations or ideas for them as they are along their journey and what their next steps might be. Oh yeah, yeah, absolutely. Actually, I'm gonna just do this very quickly. I actually prepared like a slide related to this such that the MicroMaster credential holders or the learners, if they want to know more about it. In short, the MicroMasters in supply chain management and the GCLO program, the Graduate Certificate in Logistics and Supply Chain Management, now we have this pathway that is more professionalizing, right guys? So in case that you are interested into coming to MIT during three weeks in July, three weeks in January with a capstone project in between, well, this is your call, right? So we are offering this waiver to the MicroMaster credential holders and hopefully I'm gonna be able to see a few of you or all of you in the future, right? So this is like another way to come to MIT is not the blended masters program, that's like another pathway, right? In case that you want to follow that part, this one is the GCLO program is like a glimpse of the experience that you might have at MIT. But it might help you a lot to dive in many, many concepts and to understand how to drive it into practice because our motto is actually related to how we can shape the future of supply chain management in emerging market economies, right? And I see that many of you are like living in those countries and for those who live in developed countries, maybe you are working for DHL, maybe you are working for PNG. Many of your companies are also serving emerging market economies. So it might be also interesting for you to see what you need to understand and how these countries are different from developed regions too, okay? So that would be my invitation. Thank you, Kellen. Awesome, Chris. Yeah, I'm sure many credential holders here or joining us in the future will be taking that pathway because it gives you the opportunity to actually do all this research together with Chris Mejia and his team and support your community, which is also part of what we're talking about in terms of making a change, an impact in the world. So let's launch the last poll to learn a little bit about what was the audience learning about today, what is that they found most interesting. And in the meantime, I want to bring questions from Gaving here in the audience. I hope I pronounced that well. So Gaving is saying, of course, as you mentioned, adding more constraints or more elements to a model sometimes make it either delay a little bit too longer than we need or just a little bit too complex to solve, but they are asking about what could be the sustainability controls that you could add to a model like this, just thinking on maximizing the use of cleaner transportation methods, or if you think that solving the malnutrition issue is only given like an indirect solution for sustainability. No, great question. So I didn't dive into the nutrition part, but hopefully you get it how we are handling it, right? But so to answer the question in short, so implicitly we are also reducing the empty returns of some of the babies, right? Or the trucks that we have there. However, this can be expanded into a criterion or into an objective function that we can add to the model just to make sure that we are minimizing the CO2 emissions, for instance, right? Or on the other hand, we can also pay a little bit more attention. This one is a little bit more difficult to model though to the coverage that the food that we are bringing into certain communities and this coverage, how this is being like approach, because it would be very easy for us to say like, I'm satisfying this percent of the demand based on different product categories, right? Or perhaps macronutrients, that's easy to do. But again, the granularity of the data might not be enough. That's a limitation, right? But if we really want to do like a change and a change of paradigm on how to model this, I would go for something that really, you know, like affects people and is like the employment that it generates or the effects that malnutrition produces on them, you know, unemployment, lack of productivity, et cetera. And there are so many other things, like without being so philosophical, you can also model like how you can actually balance many of these different deliveries in different population segments, right? So who is like being employed out of this? Who is like receiving the benefit out of this? So in short, coverage metrics, inequality metrics, CO2 emissions, things like that. And that could be like part of what we can do. You can also pay attention to how much you are using your facilities to, you know, like make sure that the space is being occupied properly or even better, what we are doing at the moment is accounting how much food waste is generated if this is not delivered properly and on time. That's called the perishability problem that we are modeling currently with one of my postdocs here, Mauricio Gamas. Awesome, thank you, Chris. I see a couple of hands up there. So maybe if you could put your questions there in the Q&A and we'll try to look at those in the Q&A there instead of raising your hand there. But I want to maybe build on, there's a couple of questions Ray to that specific last comment you made about perishability that seems to be a, you know, that seems to be a very significant challenge with food related research. So one of the questions was, how does the current model of research you presented, did it take into account perishability through those different grains and pulses? And then another question to build off that from Amrinda is that their comment was that in Punjab, a lot of food goes to waste because it's siloed or I guess it's stored maybe perhaps for longer than past that expiration point, if you will. So how did that model taken to account that challenge there? Yeah, this is a very good couple of questions. So number one, the perishability was considered into the model, but basically what we did is preprocessing this. So in case that some of the products were not gonna be edible for consumption, they were not shipped, right? But we didn't quantify as part of this particular project, the food waste generated out of this, okay? So I think that these answers, like both of the questions in short, but let me elaborate a little bit more on what we are doing right now with other models, not with this one, okay? So we are quantifying perishability as the remaining shelf lifetime of the products, the remaining days that a particular product has before it expires, okay? Or it perishes. Naturally, what we want to do is to make sure that this keeps, this arrives to the hands of the person as soon as possible, but something that we are inserting at the moment is like how we can expand or extend the shelf life of a particular product. We can freeze it, we can cook it, we can do something else. We can lilyophilize it, transform it into powder. So that's part of the things that are now part of the circular economy approaches that we are inserting. So that's how we are extending perishability. And last comment there, so in terms of like, how we can quantify the perishability of certain products inside the facilities, this particular model, quantify it, but I didn't show like the numbers there, but given that these are grains, most of the grains have long life, okay? So the problem there is actually the plates, okay? If you see some type of beetles, that's the part that we need to manage and that's part of the food safety approach. We, I'm not gonna elaborate too much on that because that's part of another question, but that's something that we need to control using other stuff. But in terms of quantifying this, I have a couple of current students working on that, one of them working for a big retail company. And what we have found is that for very perishable products like berries, they put away times and not necessarily the storing time is the one that affects everything. So you are not able to coordinate everything synchronously, you know, like on time, you are not gonna be able to have what is called just-in-time approaches in order to reduce the inventory, peak to zero approaches. It's impressive to think on the many extensions this research can have and the many aspects that you're touching every time you talk about any single point of it. And Teddy Barat, another one here in the audience, thinking on this extension of your future research, you talk about the possibility of reducing the number of intermediate parties in the supply chain. So they are asking whether you could consider working with someone or some organization or what is the plan in terms of not having jobs lost, like the people affected by the reduced number of intermediate parties. How do you approach such kind of impacts? Yeah, we actually, we already did it in Brazil. We approached to a, let's say, e-commerce company that was bringing fruits and vegetables to the small restaurants. And we told them, like, let's bring this to the mom and pop stores, to the Kiranas, to the, you know, like the nano stores, as we call them. And we start shipping them. So in short, the answer is we have already done it. I would love to discuss with the person who is interested in doing that because what we are doing is like, we are trying to create a direct to consumer channel without killing the other ones, right? So it depends. But particularly not every type of consumer, especially like the vulnerable population segment consumers, you know? Those are the ones who really need to have access to these low cost affordable things. And the only way to do it is like by reducing the number of coaches across the supply chain. And if you have like the community of farmers living outside the city, whatever city it is, right? We did this in Rio Grande in Brazil, that is one of the biggest states. It's the Southest state there. And we were able actually to coordinate through this company that is called SUMA app, you know? To collaborate with these farmer associations and make sure that they were delivering in small to small, small folder farmer to small retailer. And that's how we have made it happen. But I would love to, you know, like discuss more with whoever is interested because this reduces the waste, reduces the inventory. But at the same time, I need to be fair. You are also increasing the risk for the small players. So that's why it's very important to coordinate this properly with some involvement from other parties. Awesome, thank you, Chris. Well, maybe before we go to our next question if we could end our poll here, maybe we'll just take a quick look at our results from our second poll here. And the question was, you know, what was the most interesting part of today's session? I just kind of wanted to kind of get some feedback on what you got out of today's session. I'm just looking at the results here. It looks like the most popular option was gaining a new perspective on supply chain challenges and the potential to use these tools. I definitely, you know, resonate with me as well. You know, these tools that we learned and we're in academic context to see them applied to kind of a really important real world problem is awesome. And I don't know, Chris, if you have any thoughts on those poll results there. Oh, that makes sense. I'm happy to know that you are finding like ways to, you know, like use what you learned today to your daily life. I think that that's the most important goal with any of these live events. So this is great. And the fact that I see that put a seed into your knowledge about food supply chains is also gratifying and rewarding to know. Awesome, thank you. So we have two couple of minutes left here. It's made time for kind of one more quick question. This is also just kind of a question on that the current project that you presented and the research you presented, looking at the production levels. And so there's a couple of questions that I'm going to tie together here, one from Marco. His question was, was the constraint in the model, was the overall production the same or did the overall production within the region increase or decrease? That's kind of one question related to the modeling. And the second question was how the climates of the various regions were taken into account. And that to me, that sounds like the capacities within each of the regions, but maybe you'd elaborate on that a little. Yeah, yeah, yeah, no, actually the two questions are related, right? So number one, not really like the yield that is the crop capacity to keep harvesting, you know, like that you go and harvest this out of the land is different depending on the region, right? So if I go to an arid or like a desertic area, you're not going to have like the same yield, right? So that's the way that we consider it because the crop yield information that is actually collected by the government is already reflected there. We could have done more, you know, like research in terms of a scenario analysis, what would have happened if the climate change impacts these, reduces these or that. And by the way, we are currently like preparing a manuscript to submit this to an important journal. So probably if Sanchez is somewhere there, otherwise I will let her know all the feedback. That's something that we can do, okay? So now in terms of the productivity or the production, how we consider this was this a steady? No, this was different, right? Depending on the region too, because there are regions that based on their characteristics, not only now on the weather condition, but also on other topics related to the infrastructure maturity level or like the accessibility to water, the accessibility to credit certifications, things like that, they show a different productivity. So naturally this was different depending on the regions. Again, coming back to the point of the scenario analysis, probably we can play a little bit with that in order to see what would be the effective, the west or the south or the north, you know, like regions like that in India would have a more, let's say a visible impact in case that we change the productivity, the guilt or any of the characteristics, even the demands, you know? Thank you, Chris. And we are over the hour. Thanks everyone for staying with us and learning about food and retail operation lab work. We are excited to have been hosted here, Chris, and we hope to see you again. We know that there's so much more we would love to learn from you. So I'm sure the audience here is calling for you, which means that they are also interested in joining us again. Thank you very much. We're very happy to share also all the questions that remain in the Q&A feature because there are so many questions we didn't get to answer in case those are helpful for you and Sanchita and Tatiana. And yeah, thank you for joining us. Any final words to our audience you wanna share? Well, just to thank everyone for, you know, your patience to be here with us today. And yeah, I'm really happy that I provided, as we say in the team, some food for thought. Awesome, thank you. Thank you, Kellen. Always happy to co-host with you. Thank you, Lara. Pleasure and always a pleasure to co-host with you and thank you, Chris, for your time and thank you everyone for your participation today. Awesome, and everyone, stay tuned for the upcoming webinar in the series. The last one is coming soon and I hope to continue seeing you around the MicroMasters program. Thank you, everyone. Thank you, guys.