 So, I will just walk you through some of the things how we designed the machine, the micromachining center. The micromachining center what we did was, we found out what the issues were. The first one was that we do not want any vibration. So, we use the entire machine as granite, because granite has very good anti-daping properties. Then what we did was, the tool stiffness is an issue that we talked about a little bit. You can go very, very high speeds to make sure that you cut very small in one rotation. If you get very high rotational speed, you actually cut very, very little. The forces go down. And then I want very high crystalline precision. So we basically have a very good structure, optimized structure. And then I can have high torque spindle and low cost, because I take a very good Z stage and XY stage is actually precise, but not super-duper precise. When I am doing micromachining, what happens is, so if somebody wants to say the parameter of any this stiffness, damping ratio or natural frequency, you know how they get those parameters. They actually excite the structure. So, excitation is done either by a shaker to excite it or by a hammer. Hit the hammer, hit it and then take something which will measure the vibration. That can be an accelerometer, which will measure the acceleration or a displacement sensor. We did not hit right at the end, but the shank is thick. The shank is 3 mm. The bottom is 100 micron. So you do not hit the 100 micron because that will break. So what do you do? Hit at the shank. Measure at 100 micron. That is a challenge, but it has to be done. So then we were able to create all these features. These are 100 micron channels, 400 micron channels just to show what we can do with this. The other thing which we designed a machine, which we would actually just take laser, soft in the material and then machine it. So that way the stiffness can be counted because now the material is softer. So we did it in a milling setup, in a turning setup. Then we talked a little bit about this additive manufacturing. But the scientific issue with that is because when you heat it, there is differential contraction, there is microstructural changes and those induce residual stresses. So we need to understand that. And then we had nice model and everything. Now I want to build one. Now I understand the process. I have papers. Now what brings me to the most happiness is building stuff. So there was one student, good student who actually rigged the machine. He designed his own nozzle, this is a 3D printed one in the left. Then he actually machined the entire nozzle. He built his own makeshift powder delivery system. See, it's intactors, but it works. And then he printed some ring, some boxes out of it. Then we showed these small contraptions and then went to DST, gave us money. They gave us few crores. Then the real job started. Then we started building the real deal. What a big heavy-duty robot, got a real cladding head. Cladding head basically means it supplies the material and the laser, actually. So this is what we are doing. We are actually doing this for Bharat Forge. We are depositing a very heavy, heavy hard material on top of their molds. So tomorrow if they use this kind of additional layer, the mold life will increase. So it's called hard-facing. Got a powerful laser, 3 kilo or powerful laser. And then we printed something very nice and improved the machine. Now we have two stations there. One is the robot, the other is a 5-axis system. And we have designed this and got it built. So if you go ahead, you can polish it, it looked ugly, right? If you polish it, clean it up, it comes out nice. So it's 5 times normal steel strength. So if I were to cut it, I wouldn't have found a tool to cut it. It's so hard. So that's the beauty that this process can be a process where I can print super hard things, which is not possible by using machine. And then we deposited it at different angles. Because if I do it like this, I'll lose a lot of powder. Because the powder gets skewed. We're trying to understand how powder comes. Because if I have to build in 3D configurations, I need to know how the powder, how much powder actually, you call it powder catchment efficiency. And how the material flows. So these are actually cross sections after. So we did it at different angles. And if you do it really, really high angle, you actually lose most of the powder. The bottom portion, that's what it shows. What a powder is lost. So that was the actual machine part of it. Now if I want to repair it, I told you, right, we also need to scan the defect. So I'll show you another video of scanning. So this is what we will propose to do. This will sit on the robot. This is just a dummy. Easy to do. But this will be done at a real product level. There will be a line scan, which will scan the entire product. And then build models like this. So right now, I know there's a defect right in the center. These robots have accuracy of 50 microns. That's what they claim. It's much higher, the vendor claims that it is 50 micron resolution. But I'll still take it to the pinch of salt. Robots are not very precise things to do. If you want very precise, then you will have to do something what I did with the ball screw stages or linear motor stages, what I have built on the micromachining center. Those are nanometric stages. So if you want to precise, then you will have to use a different set of kinematics. Robot is not a, it's very flexible, but it doesn't have the same accuracy which you have in a gantry based system. So eventually our dream is that there will be stations. The robot will go ahead and it will pick first the scanner, scan it, place the scanner at the holder, then go pick the printing head, print it, there will be a finishing tool, post deposition will finish it, keep the thing back, then take inspection head, inspect it. So that's our final dream, totally autonomous. That's our dream. When we make one, we'll sell one. But that's our final dream that it should be totally automated. Everything is not 2D, it's a line scanner, but this line scanner is in space. So I have the data, I have the data of where these lines lie, where these lines lie in space. So this line would give me, but this line is, every line is at some XYZ of the robot. I can combine all the data. I can get the data from the robot, get data from line scanner and create a 3D map of it. So that's what a student does. The machine which I'm talking about, the vent cleaning machine, which I want to showcase today, this is how they do it in the shop floor. So he can do in less than a second, one hole he does it. We were doing 1.6 seconds, our machine was doing 1.6 seconds. We could do faster. I told him I can program my laser to move faster, but then what will happen, all my optics will go away because of the inertial forces. Because if I do very fast, there would be acceleration, right? And these acceleration will exert forces on my optics and everything. But then the drill breaks. So this is the problem that if the drill breaks, then and it goes into the tire. So you screw up one tire, but the OEMs will return everything because OEMs, if you know, I don't know how many people know how auto industry works, the only people who gets crushed is the Tier 1, Tier 2 supplier. Cars are getting cheaper by the day. And the only people whose profit go down is the supplier of the OEM. The OEM doesn't reduce its margin. So if you buy a tire in the market for 2500 rupees, the OEM buys it for 1000 rupees. That's the price. But they still sell it to OEM because they want their tires to go into Renault or Maruti. So quality of tires were not considered metal leftovers in the vent cleaning. Then drill bit can be dangerous and potentially hazardous as tires can burst, right? So that's how they walked in to the lab and said, do it. So I asked them, first thing, you just send me all these clogged bolts and I'll play with it. So we got those molds in the lab, one of the students came, he fired the lasers. Is it getting cleaned or not? So that was the first step. We said, it's getting cleaned. We saw, we observed it was getting cleaned. And they said that, okay, why don't we clean it with the air jet? So we put up an air jet and cleaned it. Got it even better. Then they said, let's try nitrogen. Try nitrogen. Is it improving? Try oxygen. Is it improving? So nitrogen, oxygen and air didn't make a much difference. So I said, okay, it doesn't matter if you don't require an inert gas to improve the efficiency. That was the fundamental idea which, so people do a lot of experimentation when they sit in the lab. And then they said that, how do we correlate how many pulses we need? So they took a tire piece, kept on firing laser on it and said, how much deep it goes? So they basically did a test by making holes, actual tire. They took a tire and they kept on fire, 10 pulses, 20 pulses, 50 pulses, different pulse energies because I needed to know what parameters are required. So this is how the tire-building machines look. This is actual factory shop. So this is the final process of making the tire. First they just make a drum and then they make the sidewall. This is the final operation where the sidewall is already made. So prototype-building what we did was, this is a two-piece mold. There will be one bottom and one piece in the top. It takes something called a green tire, which is not cured yet. It's called a pre-shaped tire is there with no treads or anything. It will come into the mold, all the treads, all the features which you see the tire comes in the curing mold. So they will take steam and they use the steam to heat it up and hold at certain temperature and the temperature tolerances are very high. The entire thing should have few degrees, one or two degrees within two or three degrees to make sure that the temperature difference is not very high. Otherwise the quality will be compromised. So they want us to do another project now. They want an induction heating-based system for curing the mold. But if I do something like induction heating, there is a gradient. So they actually want us to redesign the entire thing. So we'll set that we'll actually re-engineer the induction coil design. We'll do some physics-based modeling to figure out what would be an optimal coil design and then do a conduction study of the mold and then make sure that we design the mold with a combination of studies and some control. So these are eight-piece molds, eight-segment molds made of aluminum. So those are assembled and you see the vents will be somewhere here. All these vents are there. These are the vents which would have to be cleaned. Now the development of the test bed is initially I did not have the position accurately. So we would do the alignment by the usual deal. You see it's right at the normal, normal vector is there. So we did not have a robot or we did not have a positioning system of five axes. So we managed to do it, doing some tilting, something like that and then trying to make sure that our idea was just to see the proof of concept, whether it is cleaning or not and then get to the kinematics of it. So we did some experiment. This is the very first few experiments that you did. The first one we did with the continuous-wave laser. CW laser is not the right tool because with pulse laser, I get a lot more variables. I can do a number of cycles. I can give a lot more energy in small times. I can ablate the material very fast. So then we switched from CW to nanosecond pulse lasers and the nanosecond pulse lasers are much cheaper, 1.5 lakhs and this 20 watt is continuous power. Peak power can be 20 kilowatts. So it's actually the peak power in nanoseconds if you see because 20 watt is the average power. If I keep it on only for a few nanoseconds, 100 nanoseconds, 200 nanoseconds, the peak power is very high. It is an actual kilowatts. So that was the basic thing that the three-axis position system does not serve the purpose because I need to be normal to the surface. Whatever the surface is, I need to be normal vector to the surface. So I need a flexible position system, preferably a five or six-axis position system. So that's how we build a, we execute in a robot. So this is just the kinematics now. I need to be normal to the surface. But then the primary challenge is how do I know where the normal is? That's a bigger challenge. Now I know I can position it, I have a machine to do it and I can clean it. What do I do? How do I know where the normal lies? So a robot can do two things. I can teach a robot. The simplest would be I go ahead and teach, this is one went, this is one went, this is one went, this is one went, this is one went, this is one went. And teach one mold completely. And it records all the locations and goes and fires there. That's the easiest way to do. And the stupidest way to do. Because if I move it here, everything is gone. The position of the holes in every mold is the same, set the initial position. Ideally. So they have drilled more holes. And the thing is the moment, if I train it right, I need to either make a very good fixture or registration system to make sure that every time it knows where exactly it is, where the zero is. I move it a little bit, everything on. This is the first robot C8 got actually. Before that, they never seen a robot, right? And that's one of the bigger companies in the country. You teach the robot and then we'll give you a registration algorithm. So you'll put some markers in every mold. You get those markers, right? And then transform whatever data you had with respect to those markers. Every time you load, register the thing. A month later, they said, why don't you build us the detection system? After taking the machine, you showed them, trained them, we showed the registration thing, gave them the machine. Then they said, why don't you build it? And they said, give it in three months. I said, look, three months, nothing can be done. I can't guarantee the purchase order to build anything. So you know what? My student has a company. He will build it for you. So two months later, they come back again. Why don't you write? Why don't you build us the software for that? So we actually built a software which detects the holes automatically. It has an algorithm built into and defy where the molds are. So it's a camera system and it has also some sensors. So it knows the location. So it knows where the normal lies. This is the wind detection system. We actually built a software, a vision based software for them. And defies the holes and then goes in fire. So it records the data of the holes and then goes and fires it. So that was the second iteration. Okay, so you understand what the problem was that you need to position it in. So we used a robot for that because we could not build... Right now we have, we actually have built the FISER system. So we can use that platform for wind cleaning system. So now this is how we built it. We actually designed an enclosure, mounted the robot on that enclosure. Did all the electrical subsystems, PLC controls because the laser, the robot, the air supply, the camera, everything has to be on one platform. This is how it looked. And then we gave a nice window to do that. This is how the machine looked. Even all the electric lighting, also we did it, right? So these are the main components here. The robotic arm, the camera, the laser collimator and the compressed air jet. So there's a slight valve for the air jet also. The air jet, yeah, so air jet, so this ablates, brings out debris. This debris gets blown away. So you actually need an air jet to blow the debris up. What do you have ablated? So we use a 75 watt, pulsed fiber laser, robot, compressed laser and a camera. So this was the entire thing, entire components which were there. So this is the final, finally those guys came, we trained them. This is how the lab looked before. So typically what happens is, these lasers have 1.1 millijoule pulse energy. So the power which we'll give you is 1.1 millijoule for 50 or 118 nanosecond. So you can do the math. 1.5 millijoule divided by 150 nanosecond, what comes out the way? It should be close to 20 kilowatts or so. Give or take, if you do the math. So this can do 100 kilohertz frequency. So it can do very fast frequencies. Otherwise what happens is, if frequency is very low, then I'll not be able to do it very fast. So the pulsing frequency is very high. And then we also, we are trying to develop impressed cleaning for the mold. Now not the vent, not the deep vent cleaning, but just the mold cleaning. This is a line laser-based thing with the Galvo scanner and it will clean it. So right now it's a very cheap laser. That's why the cleaning quality is not up to satisfaction because as I said, the laser and the Galvo cost me 1.2 lakhs. The student went and bought from their own money and they build a lot of toys with that. So they clean it, comes out to be clean. So actual laser cleaning applications requires a much higher pulse. So right now I have 1.2 millijoule. I did require one joule, but 1.2 millijoule cost me 1.5 lakh rupees. One joule cost me $114,000. Do the math. $114,000 is what? 75 lakhs plus these days as a study it was like 71 or 72, I don't know. It's again skyrocketing, 71 something. So you do the math. Plus 25% loading if it comes to India. And here I just go and buy, actually it was not 1.2 lakhs. They just dragged everything with the laser controller scanner, Galvo scanner, F theta lens, which is actually doing some cleaning. So that's the thing that if you really want to do good work, it costs money, but then some compromise if you do, you can actually do it really, really cheap. That's the most of the story I want to give you. That if you're smart enough and if you understand a little bit, you can actually do a few things at a much, much cheaper price. But again, if you really want to do it good. So a German laser cleaning company says that don't contact if we have anything less than $200,000 budget. Don't even talk to us. It actually says on its website laser cleaning is not cheap. And this is what we counter. It has to be cheap. Otherwise who will buy it? So any questions? I think we are coming to close on this topic. So like there are no poisonous things. That's a good question. You actually need a scrubber here. There will be fumes. So there are systems where you can have a scrubbing mechanism that we can make. But it's a very good point. If you really want to have a good product, if I want to sell it in an European country, I need a scrubber here. So there's a very simple way to do it. What you do is you actually take a heavy gas and scrub it from the bottom. So you take a heavy gas, which will actually take all the fumes and you can extract it from the bottom. So we can actually design something like that. I have some ideas to do that. Argan. Probably Argan you can do. Whatever is slightly heavier than the nitrogen, because air is heavy. If it's heavy with that air, you can actually scrub it from the bottom. That's the typical safest way to do it. We should be very, very cautious when you design products. One of the important components is that we should also have a very strong idea about whether there is a a health hazard and a remedy hazard, both. Your personal health and environment as well. Which unfortunately I have not shown, but I think that's a component which should go into an engineering design. Thanks.