0:15 Hello everybody. I’m Rhushik Matroja. I’m one of the co-founder and CEO of Cognitive Design Systems Electric Works. We are a company based out of Kulus, not so far from here. A beautiful place if you have time 4 hours drive along Nigerian Sea. We work with large organizations in aerospace, automotive, defense and industrial applications such as SAS, Toyota etc. And while working in the industry for last 15 years, I have found and I think this is more recent challenges that we have found that the time to market is being compressed and it’s the challenges are are are increasing there are competition that is increasing from Asian side.

1:13 We have one example that in automotive industry they are looking at that targeting 1.5 year of design cycle time used to be three it used to be way more before so how to how to answer to that another one is the the talent scarcity and the knowledge loss the Janzi generation we all know we try to to to move things forward the average time has been reduced drastically that you spend in a design office.

1:48 So how do you transfer the knowhow that has been collected over the years to the new generation of engineers? Another challenge also very much European I would say is is the performance and sustainability mandates in aerospace and automotive industry. We have this net zero goals for 2050. How to tackle that and where we can we can actually use it. And now the the latest challenge is the AI urgency without clear road map.

2:20 And this is this has been something that has been in discussions since January. I interviewed 25 large organizations and everybody has okay we are investing heavily in AI but we don’t know exactly how we are going to use it and there is no clear road map for engineering itself specifically. To work on it. We came we have a we have a dream and that’s our vision is to transform engineering into what I call it cognitive creation.

2:59 I want to break the silos. I think many people before mentioned about the silos. Great talks by the way. And those silos are very much still present in large organizations and how to challenge them is is a is a is a tough tough question. If we are able to to go above that, we can accelerate the design engineering and we want to do it by doing orchestration through AI because it can bring lot of lot more value there.

3:33 So our focus is on optimizing mechanical and pro mechanical components. So we started by developing key technologies. We have our own geometry kernel based on implicit modeling. We combined it with Brap and mesh modeling. We encapsulate it. We we captured it into into a parametric design workflow that I will show you in the software. And we have also connected it with with FEA solution but also manufacturing analysis, cost analysis, carbon footprint analysis.

4:04 And all this to answer four to analyze four KPIs performance, manufacturability, sustainability and cost. And now we are going a step further where we are able to bring the AI or architect the whole workflow automatically. To do it we developed a solution called cognitive design. It’s a it’s a workflow like like Grasshopper inspired by but also with with capabilities of generative design design exploration and design automation.

4:53 So this allowed engineers to to go way further. We were able to to reach a target of like 80% of of design cycle reduction for challenges like challenges with saffron or with palace. And today I’m going to present you one of the the case study and through that case study I want I have two questions at the end of my my presentation that I would love to chat with you at the end of my after the at the break.

5:29 So the big question is how AI is changing the mechanical engineering. And let’s go through a very engineering problem that we we we had to tackle last year with a company called Palisin Space satellite manufacturer of French Italian collaboration. They had to develop 80 different variants of those small tiny tripods but they were each one of them were different. It’s for the whole con constellation. Each one of them if you try to design it manually do the simulation do business analysis it takes 30 to 40 hours on an average.

6:14 What if we can automate this? So they came to us with this challenge that can we create a single workflow where I can just change the input and rerun everything once again. So that’s what we did with our our solution. So our software looks like this where you can connect different components between them and create a workflow. So yeah, I completely agree with the person from Raven mentioning I love the interface of of note based environment.

6:46 So we we did topology optimization first. We we had some interesting results but they were not enough to to be used as it is. So what we did here is that okay like we asked we we we connected it to a component called topology optimization post process which does the the cleaning job automatically. So this kind of automation is very easy to do in this kind of environment and we have automated all those timeconuming processes on our side.

7:16 After that we perform FE analysis where we we saw that okay there was a a spike in in stress at a at a very very small zone. So we connected it to what we call simulationdriven design. Simulation driven design is basically I take the the simulation results and I optimize the design. Here we do variable offset. So based on the stress analysis that the the modification the finetuning could be done automatically without having to do the CAD operations and then we brought it to manufacturing driven design.

7:52 That’s something that we really care about is how to make the part manufacturable and when everything is done once again we validate the results. Perfect. Everything is great. U the results looks much better than before. We have a part which is lightweighted by almost 92% something like that. And then we did manufacturing cost and carbon footprint analysis for for a a specific manufacturing process material combination that we have and that we launched it to to our reporting tool which is a design exploration.

8:33 So here I can vary the parameters previously on all my workflow and I can create different variants. I can automate it also in a way. So we can create lots of different variants and do the comparative analysis on all different KPIs that I mentioned earlier performance manufacturability cost and carbon footprint. When you are satisfied with one of the the manu the the process then you can bring it back into the into the into the workflow modify the the initial design and rerun the the whole workflow and it will never break.

9:12 That’s the promise that we are we are giving here. So what we did here is that we we we created a workflow which is a very typical engineering workflow for lightweight applications that we have. We did design exploration. We identify which workflow works the best so we can standardize the design process itself. Once we have standardized the design process, we extracted that workflow and said okay now do the same thing for other parts.

9:50 So the whole design process became much more fluid and accelerated. So this was done in September 2025. But then everything changed with AI. So we asked the same question to AI. Okay, what can be done if I connect all my technology through an MCP server and instead of bringing a chatbot into my environment, which is not a bad idea, but we we do other way around, I bring my technology to a more widely commercially available solutions which are evolving at a at a speed incredible speed.

10:33 So now we can just ask here in this case we asked to do cloud I’m sure many of you guys are using perform optimization on the structural bracket which is in this folder read the specifications read the the 3D models and and go for it AI did the the project orchestration it did the workflow management it analyzed the KPIs and it created the next step also for it and it gave some other ideas that were not always we didn’t think about it.

11:08 So it was interesting but the execution was done on premises. So we did it on our PC. So all the data that was created simulation data everything remained on a local environment and this is something which is important for our customer. They don’t yet like that much to to put everything on AWS. Nothing against them. But yeah, that’s the that’s the story of of our current customer.

11:34 And it generated the similar kind of design and the report very easily. Let’s go through it in in couple of seconds. So this is the specification I gave very very detailed like you see and cloth start running on it. It orchestrated v different steps. There are 13 different steps. It created a design. Okay, that’s not great. Can you optimize it further more for manufacturability? Do post-process on the topology.

It it did the the post-process also. Okay, now it looks much better. Do the the FE analysis. Go for it. Yeah. So, it ran the the FE analysis. It it passed the the the criteria and created a beautiful report. Things that took that took Talis 100 hours. With cognitive design, we were able to reduce it already to sorry 600 hour to to 100 hours. The whole project and now with with GL it was in in hours or minutes the this whole change in and the in the dynamics were very important and we started asking questions to our our our users like what’s the new role of of a design engineer and where we are heading to and what’s the right way to do it.

13:13 Sorry. There you go. First the the most important thing for them that came out is that we are working in a very constrained environment where certification is necessary. We need deterministic workflows. We need workflows that one sets of input has to be one set of output. So I can create cibility. I can I can have certification ready workflows. If something happens, we should be able to know why it happened and how can we improve and for that creating an infrastructure where the deterministic workflows like FEA design optimization etc could be orchestrated by AI is important but also it’s also important to capture the patterns the that we can have in our engineering work.

14:14 So we do sim if if the similar workflow has been presented again they do the similar job and not comes up with a completely different workflow. So this is the the technology that is being being developed. But this is bringing to a big question. The big paradigm shift is happening. Role of a design engineer is not going to to disappear but it’s going to not transform and this is this is I would like to know your opinion on it.

14:47 How do you think it is is going to transform? I believe that the modeling and drawing will become design intent definition like John before mentioned about it like okay give your design intent and the AI will do the work running iterations becomes orchestrating AI pipelines completely agree on that and checking the results will become certifying AI outputs. So engineers who used to like so the whole generation of engineers who grew up with with all this CAD systems they know where to find the right commands in which menu and they are expert in in knowing a CAD software but what we need is not an expert in a CAD software.

15:35 What we need an expert in mechanical engineering and the big difference will be happening in the the coming coming years where we have the the operators and the the the engineers will be very much differentiated thanks to AI and kind of in a way we are going back to how engineering used to work where engineer had their sense of understanding if the result is provided to me is it the right result stress is correct displacement is correct or not etc.

16:21 So our goal is to to bring the determinism in the the the real world engineering and we started working with various different organizations like in France and in Japan mainly at the moment and also in Italy. We are bringing concurrent engineering to to the reality where different various different roles could work together collaborate and accelerate the design process. By using the deterministic approach keeping the human in the loop.

16:52 So the the the pipelines that has been generated are validated and remains in full control by the human and it brings AI to the regulated industry. While coming in coming here I was asking myself a few question that AI is bringing fluidity of interface between the human and the technology the the barrier that we had and it yeah I call it barrier but yeah the interface that we have looks like a software with a software is not allowing us always to communicate better the design intent to the technology chatbot is one of the the way of doing this interface is it the right one wrong one I’m not sure I’m not convinced but happy to discuss about it but neither is the monolithic interface that we can have in a card environment you can’t expect to design a washing machine and an aircraft engine using same card interface it doesn’t make any sense.

18:08 Now we have possibility to create an interface which is suitable for a specific requirements and we should be able to to customize it for a specific industry specific customer that would be very very important and I would love to discuss also on the on the curated AI that we are bringing into CAD like what’s the right way to do bringing AI into a into a software or bringing the technology to an AI interface which is much more flexible.

18:44 I think there is no right or wrong answers into it but happy to discuss. Thank you for listening to me. See you around. To learn more about the CDFM computational design symposium, access the archive of previous presentations, interviews with speakers, and information about future events around the world. Visit CDFAM.com.