Training AI Algorithms for Automation using #NoCode.
Harshal Trivedi is the founder of 3 successful IT Ventures with 15 years of experience, an Artificial Intelligence enthusiast, and currently focused on scaling Tusker AI.
Abdulaziz M Alhamdan 0:07
Once upon a time, there were millions of businesses struggling. Every day they wasted time, effort and money on the repetitive tasks that added no value one day, the better automation podcast by process CEO came to help them find a way. Because of this, these businesses save time, reduce costs, innovate and make better decisions because of that. These businesses grow, scale and use human creativity to change this world. Hello, my name is Aziz and I'm your host that better automation podcast by process CO, where I interview the world's top experts and share their very best ideas on how to improve automation in your business, processes and life. My guest today is Harshal Trivedi. Harshal is the founder of Tusker AI, an enterprise grade automation toolkit blended of no code AI platforms for all business communities that want to run things within a fraction of the cost. Harshal is the founder of three successful it ventures with 15 years of experience. He is an artificial intelligence enthusiast, and currently focused on developing a no code computer vision platform. Harshal How are you today?
Harshal Trivedi 1:45
Abdulaziz M Alhamdan 1:46
I'm happy to have you here. I'm excited. And I want for you to present the story of Tusker AI , how did you find the need for it? What is? Why did you find that now is the right time to move into the AI field and how to use it for automation? And what difference does it make?
Harshal Trivedi 2:10
To start with like I was into automation from last like 10 to 20 years, I started my automation journey from the early days of my college where like I developed a Braille hardware for the blind people. So I supposed to visit a blind school over here, which is the ACS biggest blind school. And what I saw was like people were able to, like use computers, they were able to use their phones. But as an engineer, what I saw was they were not able to read high end technical stuffs, the books, which were there were only till the school levels. So they are well. It's part my journey to automation, where we developed a hardware from where they can read the high end engineering books from the real hardware. So they're from the early days, that spark, then I was associated with one, it was like a contributor to one of the early stage auto ml projects started by Carnegie Mellon professors. So I was only a contributor there. And later, it was acquired by Apple. So that made my journey into AI. So that was the first stepping stone where I started my journey into the auto ml space. Now talking about Tasker and how it started. So I'm in this particular space from last like five to seven years. And what I see over the years, talking to multiple customers, like customers spends millions of dollars in automating their processes into AI. They need to have large technical teams to create AI models. And what happens is it takes like months and months of time to create one or two models. And after spending months of time, and capital. Still, there are multiple projects that fail into POCs or MVP, they don't go into production. So there is lot of black box for the stakeholders when it comes to AI. So if the models are made, they are not optimized enough to go to the production or the production cost is too heavy. Or they'll come to know the data is not proper, or data is insufficient, or only the part of the data is useful. So that's part why not to create a tool that can just use three simple clicks, upload, train, and deploy in three simple clicks. Without writing the code. And without having the knowledge of AI. You can build your own AI models and deploy into production in just a few hours. So this is the overall journey. But it's part to start Tusker.
Abdulaziz M Alhamdan 5:05
Thank you. So if I understood you correctly, you know, for many companies, it's a big investment, whenever they need to create an AI model, they need a team of highly skilled people. And even then, they might find out that it's not working as it should. So they cannot use it in real life. Or there is a problem with how it handles the data, or how the data, whether there is enough data and all that and all that money was lost and time. And so what Tusker does is three steps, where they will upload whatever they need to upload, they will train the AI algorithm or model. And when it's ready, they will deploy it. And therefore, it becomes a matter of hours where they can iterate, they can test and they can find out the results at a fraction of the cost, as well as a much, much faster turned around and turnover over time. Is that correct?
Harshal Trivedi 6:02
Abdulaziz M Alhamdan 6:03
Thank you. Well, how is that related to automation? Because we spoke about AI, about the AI models, and you are very versed in automation? How can people interested in automation, use something like Tasker or an AI optimized algorithm or something like that, in order to improve their automation or data reporting, or test the quality of the data available and find trends they didn't see in the data from before or tell me what would be some case studies or use cases that you find useful?
Harshal Trivedi 6:43
Taking on the customer perspective. So as we see, they all have the infrastructure in place, as they are like businesses running from like, couple of years. So they have tons and tons of data, in terms of if we take image and video data. So there are terabytes and terabytes of data, which is stored every month. But the thing is, the complete processes are manual as of now. And businesses are not educated enough to have to take actionable insights from the data which they have now using Tasker, and it's a no vote. So it can work in multiple domains, we'll just take a simple example, into a manufacturing domain, where manufacturing plant can automate their SOPs, standard operating procedures, their anomalies, their compliance, and their security requirements, and others with their CCTV data infrastructure, which they already have in place, in just hours of time, all in just few days, giving a thing, simple examples into SOPs. In manufacturing, there are multiple standard operating processes which need to be followed. So, suppose if a worker is working inside a manufacturing plant, there are a lot of SOPs that he needs to wear the red hat hardhat, he needs to wear the vest jacket, and he needs to wear the gloves. These are simple compliance part. But isn't when there is a problem, it breaks the complete production pipeline, there is a chaos, and there would be hours of loss into the production of the manufacturing plant. So through Tasker, they just need to upload this particular data, uploading the data on a single click, they can train the model and deploy that's it within a day, complete SOP for the PPE compliance personal privacy equipment compliance will get automated. So, this was one of the examples. Other if we look at the anomalies, then there are particular spaces where we work with one of the largest copper factory where the scrap of the copper and the actual copper which is made in the cables, the cost of both the sin. So, there are a lot of theft issues month on month happening in that particular manufacturing plant. So, for that, in terms of anomaly, if they have some restricted area, where the scrap is there, if they can authorize few persons to only enter that, then through a simple CCTV camera, we can automate the complete process where real time alerting systems can be generated. So these are only two examples for only a specific domain. But it can be in multiple domains and multiple use cases can be measured through your real time CCTV fields. So this is where the vision component comes in. Is this correct? So instead of only creating the process or as you said SOP. You have some visual confirmation that it was followed. And if it's not followed after you know the training that confirms that it checks for the right things, it will alert you that the SOP is not followed correctly. And therefore, you will take corrective action or what is exactly the part about the training and the AI compared to using any other simple manual SOP confirmation or SOP clarification process. So now, where there is a huge implementation, so if we take large manufacturing plants, they are across acres of land. So, at a minimum, they have more than 100 200 300 cameras on their premises. Now, manually monitoring that 300 cameras is practically not possible for a person or two, to individually monitor every fifth 24/7 It's a very hectic task. Plus, as it is manual. And as there is a person intervention in that there can be lot of things which can be missed. But through Tusker, it will only give the actionable insights where the anomalies or the SOPs were not followed. So he needs to see only the data. So it doesn't need to surveillance, the all the 24 cross seven data just needs to look at that five minutes, 10 minutes or 15 minutes of data throughout the day, which generates the real time alerts that in this particular time, and in this particular moment, in this particular area. Our SOPs were not properly followed, it reminds me of management by exception, which is that managers should only intervene if within a range that is acceptable, if the results or the outcome or anything like that is too low or way higher than expected. And only then they should intervene either to find out why things are better than they should or I why things are less, or, you know, worse or less optimal or less good than they should, which saves them a lot of time. Tell me this, and I know you're from India, and you have hundreds of millions or of people who are in the job market and the job space and all that there are some people who say, automation or something like Tasker would be reducing their job prospects or opportunities. Well, what do you see in the future, as the role of human beings to add value that AI or automation or Tasker cannot, you know, do that human beings should focus on in order to create abundance and great things in this world? So to answer this, through AAA, what happens is, many people attend to say that they will lose lots of jobs, no, but through AI, what we are focusing in what I see as a vision, it increases the productivity of a specific plant, or a specific enterprise, where it is implemented. It's not meant to track a person or lose a job of a person, but it is just to meant to make the process in such a way where it is more productive. That is the first case. In the second case, what AI will do all the static, and the manual processes which are happening, it can automate all that it can generate real time alerts, analytics, actionable items, through machine learning, you can have predictive analytics as well. So what will happen, the state task will get very much reduced for the business owners, where they're doing the businesses at scale. And we're doing they're doing in route of multiple countries and multiple setups, and there are 1000s and 1000s of employees working over there is easy to crack and easy to manage manage the process. So all the static tasks will ought will get automated. And the persons who are doing the static jobs will get more productive and do more productive tasks where they can be highly paid.
Abdulaziz M Alhamdan 14:30
Actually, I'll add even more because when a company is scaling, their biggest breaking point or issue is hiring. And therefore, by doing this, they need to hire less people. They will need to focus on hiring the right unique talent rather than hiring supervisors and people who will be doing jobs to check that the SOPs are followed and all that while AI can just take care of that do it reliably 24/7 Without to brake and saving the resources of the company to actually scale and get the key talent that will make a real difference to their bottom line, and to their results and outcomes. Then to ask you more, you spoke and gave examples that are more related to mining, manufacturing such businesses, how can other businesses take advantage of AI enabled vision focused automation like Tasker AI to make a difference in their workflow.
Harshal Trivedi 15:37
So as if we look at the generation right now, right, then everyone is majorly working over images, and videos, the complete social media if we see, it's all into the image and the video stuff, if we see everywhere, everywhere, every government, every other businesses are implementing video surveillance and other stops to make analysis of whatever data they already have. So, if you take into not manufacturing, if you take into retail, then in retail, there are multiple things where they can check the demographics means what type of customers are coming, then whether they are male or female, they can help you build a great customer experience. So if there is a visual intelligence on top of the person who is a representative to the end customer, he can check his emotions, and the business or other stakeholder can get real time insights, whether the behavior of a person with the end customers are proper or not, in the retail Other examples are, what are the areas where there are maximum. So customers are coming where customers are not coming more. So all the things can be automated into retail. Other than that, if we go into, say, for example, into the in the government space where we have worked into beaches, then it can help in saving lives of the people also. So we can do a beach surveillance to visit intelligence through a drone, where if there is a yellow mark or a red mark, where a person should not swim beyond that, if a person goes beyond that an automatic alert, get can get generated. A Watchguard can get an automatic real time notification. And you can just go and see if more tide is coming. At that point in time through vision intelligence, we can automate that. And we can evacuate the beach as early as possible. So there might be less casualties. So these are other real time use cases where vision intelligence can use. And there are much more examples than this.
Abdulaziz M Alhamdan 18:00
That sounds really, really fascinating. And since you're in touch with the state of technology nowadays and automation and all that, what trends do you see will be future trends when it comes to AI to the usage of vision. And you said everything now is more focused on pictures on videos that will either make businesses work better function more effectively, or even open new ideas and new doors for new companies and new solutions that can change the future? Do you have an idea of where the trends are going right now when it comes to AI automation and AI vision focused applications.
Harshal Trivedi 18:48
So looking at the future trends into AI, so he does go back two decades. So if you go back two decades, then there was a y2k era, which came in 2000s, where every businesses wanted to have their presence online. So they created the websites. And once it was done, then came the part. Everyone wanted it processes into ERPs, or supply chains. There were everyone implemented the supply chains and ERP. After that, the mobile era came where every businesses came into the mobile apps of Android and iOS. When every businesses go there on mobile, then there was a huge internet traffic, which was going across the world. So there was a scalability issue. And all the application and businesses were becoming global. There were cloud came in place, and make make all the businesses global, and scalable. And now sitting here right now, if we look so next trend, if we see is into data, where they have all the data in place, they've altered it As aggregator like aggregated over cloud, everyone has their online presence. So the next era would be generating more analytics out of that data. Second, getting more actionable insights into getting predictive analytics. Third will be more and more image and video surveillance is used. So that will pick up the market in the next decade. And for the future, if we see like, voice, image, and video will rule the next decade into AI space.
Abdulaziz M Alhamdan 20:44
Yes, I am really excited about the future. I find it fascinating all the opportunities that are open all the lives that will be truly changed all the extra value that will be there. And I'm sure that Tasker will be a big part of this. Can you speak a bit more about Tasker? If people want to use it? Or to learn more about it? Where should they go? Can you just summarize the value proposition in a few sentences, and I will make sure to write the website and the description as well.
Harshal Trivedi 21:18
So to use tusker, what we are aiming is, is into the touchless acquisition into image and video analytics, in our overall vision is to democratizing computer vision, at scale, where anyone, and everyone can come and build their AI models, without having to learn AI. To reach out to us, you can visit our website www.koco.ai. You can go there, you can request for a register and request for support, where our support staff will get in touch touch with you. And we'll bring you on board. We'll understand your needs and will explain you educate you that how the data of images and video which you have can be generated to more insights can generate more value to your business, generate more ROI to your business. And in less than like two to three days. With Tasker, you can easily start your automation journey.
Abdulaziz M Alhamdan 22:33
Thank you i It is very, very useful. And on my side. The tool I'm impressed by and it's making this whole podcast possible is PROCESIO you that I recommend to everybody. It is the modern low code, no code platform for advanced automation, and creating an enterprise grade back end for your software that I'm sure it can integrate very nicely with Tusker AI for even more power. Any listener can get free access at processor dot app for a free account they can use and when they're ready to upgrade and our fans and want to use it even more. There is a very generous discount code. It's called BETTER50OFF, one word in capital letters. It's written here, and you will get 50% discount on upgrades. You can find more info in the description. Thank you Harshal. It was my pleasure, my honor. And I wish you to keep going because you're doing good things. Thank you so much.
Harshal Trivedi 23:39
Thank you for hosting me. And thank you for like for the podcast and all and thank you everyone for listening. And thank you again for hosting. Thank you