The AI Guy

Steve Melito: Hey everybody, welcome to New York State Manufacturing Now, the podcast that’s powered by FuzeHub. I’m your host, Steve Melito. Today we’re talking to Anoop Bhatia, the CEO and founder of Nowigence, incorporated in Albany, new York. Nowigence is developing artificial intelligence solutions, and AI in manufacturing is what we’ll talk about today. Anoop, welcome to New York State Manufacturing Now.

Anoop Bhatia: Thanks a lot, Steve. A pleasure to meet you and your audience. Thanks for inviting me over.

Steve Melito: You’re most welcome, so please tell us about yourself. Your LinkedIn profile indicates that you have a degree in chemical engineering and worked for GE and then Momentum Performance Materials for over two decades, and then you started Nowigence in 2016.

Anoop Bhatia: Right. I was hired by GenElectric in India way back in the mid-90s and at that time we were very young in globalization. We knew the spelling, but we didn’t know the problems associated with globalization. During that era, information was very tough to get. So when a company starts globalizing and starts understanding the markets in different regions, there’s language, there’s currency barriers, there are people barriers. But that is how I got involved with GE. I got transferred from India to Europe, was there for about eight years and then got transferred again through an acquisition to the US, and by the time that we got to 2016, which is about two decades later, at least, the problem that we saw was that there’s so much of information that decision making becomes very tough, and that’s what I think gives birth to AI and most of the companies, whether the big ones or small ones like us. We started around that period of time because at that moment, technology had matured, to kind of take it forward.

Steve Melito: Very good. And before we talk about Nowigence, I would like to talk about your company. I’d like to talk about AI, because I think it’s a term that most manufacturers have heard, but that some may not truly understand. After all, it’s not like a robot that you can put on your production line and see every day. What do manufacturers need to know about AI?

Anoop Bhatia: I think the most important part is that manufacturing has a lot of data and AI is nothing but data analytics. The change from, say, the previous world of data analytics that we lived in is the fact that now AI can contextualize textual data, and that’s where you see. You know, chad GPT was amongst an example that said that, hey, I can actually create text for you because they train the models to do that. So when you look at data analytics, you’ve got the numerical science, which is just 10% of the data that we actually consume, but the wealth of our knowledge lies in textual data, some of the images that kind of represent that. So that is where AI can help any manufacturing company, enterprises, small businesses, where you look at data analytics for different types of data, it’s called as unified data, and that unified data, when you look in its completeness, enhances critical thinking and that’s what leads to the next level of productivity or research.

Steve Melito: Very good. So is it true that with data and with AI, if it’s garbage in, it’s garbage out? In other words, in order for AI to be effective, do you have to have quality data?

Anoop Bhatia: In other words, in order for AI to be effective, do you have to have quality data? Well, a lot of your manuscripts, a lot of your logbooks and I’m talking pertaining to the manufacturing industry a lot of the papers that are written internally, they do not have garbage. That’s the wealth of knowledge and our focus is to kind of bring that for digital discovery. And our focus is to kind of bring that for digital discovery and that’s where our products basically kind of upload all that information and then it becomes accessible for better process control of a better discovery of what the root causes to solve problems in manufacturing is. But going back to your earlier point, yes, at this moment the world may be confused because most of the AI that they’ve seen is in the web research area. So when you go into web research, we know that the web is biased. We know that it is biased because of the fact that most of these, whether it’s Microsoft or Google, they earn money out of SEO and out of ranking your opinions on top of others, where you have to pay to do it. So you can always inject bias and if you’re just kind of looking at the worldwide web to provide you some ideas, they would be biased.

Steve Melito: Sure and I liked your example about good information of a logbook and making it more accessible. It’s not sitting on some supervisor’s shelf somewhere, it’s actually available to everyone. What are some other real-world applications? In other words, how can AI be used in manufacturing in other ways?

Anoop Bhatia: Right, I think I’ll respond at two levels. One at the highest level, which is what drives, I would say, technological advancements. So when you look at the world and we know that, way back from 1970s onwards, we in the US have been having declining birth rates and that pressure is coming onto the labor force. We’ve been maintaining a GDP. Gdp growth is, you know, for the world’s largest economy, at two and a half to three percent remarkable growth. But where are the people going to come in the future? And at that point you have to look as to how do you inject productivity but increase efficiency. Just productivity by itself doesn’t make a difference, but then when you have productivity with efficiency, it makes a very large difference. So that is what should be the first motivating factor when it comes to AI and even in manufacturing. The second aspect of AI is what can it do? If you were just with data analytics on, say, doing process control from just some of the few gauges that you are trying to understand, you know where your think tank from within, the people are kind of looking at that data. It takes time for us to read, absorb it and to kind of sell it internally within businesses in order to execute on them, whereas if you have AI, you do bring all that information forward. Whereas if you have AI, you do bring all that information forward, you democratize it to a certain extent, but it pulls out the root causes on which you can put your control parameters and guide your research, your innovation or your manufacturing process at a fraction of a cost and especially when people are not available. So that’s just a very again, it’s a high-level case. You can break it down to smaller use cases. I’ve got manuscripts. How do I digitize them? You know the product basically pulls out that information and gives you renders, it for digital discovery. You can ask questions, you can train your people as they are coming in. They’ve got the SOPs on safety, on manufacturing, what’s critical? How do you troubleshoot different events? It kind of helps you to save costs and drive efficiency at a pace which is unbiased and it’s controlled by you, but at the same time it reduces your pressure on just routine manual processes that otherwise you would have had to spend.

Steve Melito: Excellent. You made some great points about workforce and the cost savings. Let’s talk about smart manufacturing and AI a little bit. It’s my understanding AI is an important part of smart manufacturing, but I can tell you what I’ve seen is many small to medium enterprises have been slow to adopt smart manufacturing. But I can tell you what I’ve seen is many small to medium enterprises have been slow to adopt smart manufacturing technologies. Sometimes it’s a matter of cost up front. Sometimes it’s a matter of having the resources to implement. What will drive the adoption of AI, then? Is it really about the workforce issue, or is there something else that will get small to medium enterprises to adopt this, where they have perhaps been more resistant to things like robotics or industry 4.0 sensors?

Anoop Bhatia: Right, I mean, that’s a great question, Steve. I look at it and I say just the awareness you know, every time when we look at a problem just knowing where to go, whom to talk to that awareness itself becomes a motivator for you to bring any technology to its practical implementation level. Company, if I don’t start today, then naturally I would lose on to organizations that have accepted AI. And again, ai starts with digital discovery, which is what the web search engines are saying, and to implement digital discovery is not really costly or time-consuming. We have a product, we can render it for digital discovery and then we can take it forward and break it into steps. So the starting point is not difficult. If you look at all the information that’s going from an adoption perspective, people are very confused. You know, if you’ve got generative AI and it creates an essay for me, if I ask the same question as you would ask, then how would it differentiate us? But that’s not actually the point of Gen AI, even if you look at, we are very focused on extractive AI and we do customize some of the large language models because we’ve got a process which says, hey, I first need to understand the root causes before I can solve the problem and that’s what we try discovering. But overall, you know, just generating essays without any depth of thinking is not going to lead us to a monumental change, and that’s what we got to understand. There’s a whole lot of hype. That confuses people also because they say, hey, just Chad GPT alone is not going to help me drive all the businesses. But there is a lot of training. There’s a lot of I mean to the fact that they’ve been able to simplify that, once you have the depth of your knowledge, you can use that to create your reports, to create your dissertations. I mean, it’s a great technology and we should look at it, but don’t get caught by hype, you know. Don’t get caught by what’s happening. Talk to people who actually, you know, have a genuine interest in kind of promoting this technology. They’re researchers, they’re AI, data scientists who are actually giving their life to help businesses succeed, and we need it. In New York State, 40% of the GDP is coming from small manufacturing or small businesses, and if we don’t support or if we don’t find adoption rate increasing over there, then it’s a problem. There’s a miscommunication, there’s a communication gap occurring between two agencies and you know, people like you need to promote us also, then You’re absolutely right, and getting past the hype can be challenging.

Steve Melito: In some ways, AI reminds me of where 3D printing was about. 10 years ago You’d read articles about some hobbyist using a 3D printer to make a coffee cup. Okay, that’s kind of interesting, but if you’re a manufacturer, you look at that and say, well, that’s not me. So it’s very important for folks to understand chat. Gpt is not the extent of AI. You did mention a term large language model. Could you elaborate on what those are?

Anoop Bhatia: So the large language models have been trained with, let’s say, the world web data that’s available. So you know basically what it means is. It’s a predictive way and I’m explaining it in very simple terms here, Steve, and it’s basically you’ve seen it also that a lot of the tools that we use when we start writing a sentence, large language models at the back end are trying to predict what the next word would be. And so, when they have been trained on a lot of data, they’re trying to complete your sentence from whatever is the depth available to them in terms of the way they are extracting. And we do not know where the large language models extract. Do they extract from the last 60 days of data or do they extract from the last 500 years of data? But they kind of try to complete and there is a lot of data science behind it what’s the main thought that they would bring forward in the direction of thought that you would to predict what your you know next course of action would be. And again, it’s a lot of data analytics and these large language models have you know, when I look at the world of natural language processing, which is when we started, and we started with extracting. You know, the research group that we were involved with were successful in extracting content from doctors’ notes and creating a table at the back end so that we could do statistical analysis. And that’s needed for process control, continuous improvement, for a lot of different things, for trying to find the most productive route from symptoms to cure in a medical industry also. So you look at it and you say that was the simplest use. But NLP natural language processing requires a lot of manual effort in terms of creating that training data set. So when you look at LLMs, they take away that burden, they reduce the cost of implementing AI. So there are many benefits. Again, there are simpler ways to communicate it. There are multiple ways to look at the backend, because now, when you see large language models, they’re getting very, very advanced in trying to give you the tools so that if you connect the products at the backend correctly, you will be able to implement AI at a much reduced cost and at a better efficiency.

Steve Melito: Excellent. And the costs again, it is an obstacle for small to medium manufacturers Really leads me to my next question Is AI only for big companies or is it for any type of enterprise?

Anoop Bhatia: It’s for any type. According to me, you know you’ve got a starting point. The starting point for you is you need data for data analytics. And how are you organized with your data? And how are you organized with your data? If you can unify your data of textual data, number-based data, image-based data, that itself solves 50% of your problem, or that itself solves the speed by which you’re going to solve problem within your businesses, and it doesn’t take too much of cost and time to do that. Just make data available for discovery. That’s use case number one.

Steve Melito: Okay, If I can ask for maybe a manufacturing example. Let’s say that I’m an engineer and I need to pick an adhesive and I’ve got a bunch of different spec sheets. I could go through and look at them one by one to see which one has the certain property I need. Could AI and a large language model help me with this and save me some time?

Anoop Bhatia:So the way I would go about that problem is that there are research papers written from all around the world as far as adhesives and their stickiness to various substances are concerned. So that’s your first and you have your own internal research that you have done, for which you have created the current product stream that you’re offering to the market on Adesys, and you’ve got your customer reviews, which says that this is what works well and this is what should happen also with the product. Then I can use it elsewhere. Now all of this is there, available for humans to kind of interpret at the rate at which we read, which is 300 words per minute. You upload it into the product. It reads, you know, at the speed of what machines can read. Now you’ve got your root causes on. You know the product. At the end, I upload documents today, tomorrow, the product gives me a list of these 50 critical factors which are there in the design of these adhesives. They need some resins, they need some substances that make them stick, they need to be resistant to some atmospheric condition. Whatever be those parameters, now a business, a small business, cannot run after these 50 causal effects. We know the causes. The product, our product research work AI actually gives the causes the risk and opportunity associated with it. You choose the five that you think is going to give you the best return for the money you invest, and on that you need your control parameters. This is where your manufacturing Parameters is today. This is where your research papers are saying that. This is where your manufacturing should be at. You know there’s a delta we’re working with a mine where, even despite hiring a huge data analytical team, their yields have kind of got saturated at 88%, and every 1% increase in yield for one of them gives them $6 million a month. So now how do they solve the problem? They use our product. They basically look at the causes. They put their control parameters, their tolerance levels, to control those parameters and the yield continues to improve. No bias, just control parameters and day by day, as people understand what the causes that could have been improved by themselves, coming out from their logbooks. You know awareness is the best way to solve a problem. They get an alert, they are aware of what they did and they control the problem. So that is just one use case in the adhesive manufacturing. Also, what are your root causes? What are the parameters that you would like to control. Those are human decisions. Put it into a process. Let people get educated on those hypotheses that you have built and let the process improve.

Steve Melito: Excellent, and I think you began telling us a little bit about how your company can help, and I’m certainly grateful that you’ve talked about AI so generally. How does Nowigences fit into the AI universe and what are some ways that you can help manufacturers, regardless of their size? Let’s say they’re a small to medium manufacturer that does machining or whatever. How do you work with them so?

Anoop Bhatia: I think we you know most of the companies which are out there in AI they come with a whole bundle of consultants and service engineers and that makes it a little costly for small businesses to kind of start with. They cannot afford the infrastructure, nor the people cost, nor and data scientists are expensive. So the way that we looked at it, Steve and it took us a long time was we bundled it into a product and the product at the very beginning, I would say at very affordable prices, at a fraction of a person, would help you to kind of decide what should be the five or six causes that you would identify to kind of improve business yields. Could be in manufacturing, could be in service-oriented businesses, just a way of looking at yields. The measurement system changes. But you know, obviously every business is trying to control on certain factors that it can control. So essentially that’s the process. We can give right away, can give right away, and then, of course, with the smart manufacturing system that you mentioned, we can then customize it to bring in your data on a day-to-day basis on whatever be that frequency, and then we can set control parameters, we can set alerting, which comes later on and I think, as your business grows. You would need it because you need you know. Where are the people going to come? We’re already suffering in a generation where we do not find people. And then the question comes as to 10 years from now. You know, if you don’t start today, how are you going to manage? And it may not be 10 years, it could be just four or five years. But if you want growth, and you want growth with productivity, then you have to start with AI today, and we are there helping you out.

Steve Melito: Good and if I understand your value proposition correctly, you offer AI as software as a service, s-a-a-s. I always think of software as a product like Microsoft Word or Excel, so would you be so kind to explain what software as a service is?

Anoop Bhatia: It’s just, you know, in technology you offer a product and in technology, always there is a starting product which says hey, get going with this and in that, get going. You know, upload your documents, build your discovery system, interact with the system and then when an IT product needs to kind of measure something in pounds instead of, you know, square inches, then you do need to customize it. And if you have to put, if you have to connect your data which is coming from your wing scale or your pressure sensors, in order to you know you’ve got the data in your logbooks. It’s not that you don’t have your data, but if you want to kind of use AI to discover the insights and build up correlations without any bias of human interpretation and you want to kind of sell it quickly in the organization, from an implementation perspective, it’s always important to take a third opinion and that is what our product as a SaaS solution does. It gives you that third opinion. You can still debate. We all have to understand, Steve, ai is not to be scared of. If you look at our brains, we use five senses in order to interpret a situation and there is no power on Earth, there’s no computational capabilities on Earth today that can process those five senses all together, like our human brains does. So to say that AI is going to become more powerful today than humans. And be scared of it. Don’t get scared. It’s just a technology to help you gain efficiency at a lower cost. But if you don’t use it again, I urge you got to start using it today, because our economic conditions with so much of information, you know, declining birth rates If you don’t start using it today, you’re affecting your growth. You’re affecting your country’s growth and ai got born in this country, so use the benefits of it this has been very helpful.

Steve Melito: I’ve learned so much by talking to you today. How does the manufacturer contact you to learn more about knowledge and how you might be able to help them?

Anoop Bhatia: Oh, just send us an email. We’ll get back to you very quickly. The email is info@Nowigence and I’ll spell it out. It’s I-N-F-O at N-O-W-I-G-E-N-C-E.com. The word Nowigence is made from now and intelligence, and that’s how the name of the company got created.

Steve Melito: Excellent. Anoop Bhatia, thank you so much for being part of New York State Manufacturing Now. Thank you so much.

Anoop Bhatia: Steve for inviting me again, and a good day to you and your audience.

Steve Melito: You bet. Thank you so much, Steve, for inviting me again, and a good day to you and your audience. We’ve been talking to Anoop Bhatia, the CEO and founder of Nowigence. Incorporated in Albany, New York, Nowigence is developing artificial intelligence solutions and we hope you’ve learned more about AI and manufacturing today. Do you have questions about technologies like AI, machine learning, data analytics and robotics? FuzeHub can connect you to experts across public, private and university sectors. To get started, just let us know what you need. Go to FuzeHub.com and click the speak to an expert button. It’s right on the homepage. Then fill out the short online form and let us know what you’re looking for. A member of our Manufacturing Solutions Program will be in touch within 24 hours or the next business day, on behalf of FuzeHub and New York State Manufacturing Now, this is Steve Melito signing off.