0:28 So, my name’s TJ Root, and I’m here to talk about InfinitForm, which is a unified unified design space for the physical world. For the last 30-odd years, the best design is rarely been made, right? What gets made is what gets through the process. The process becomes the constraint. What’s that process, right? In a perfect ideal world, something like this. It flows, right? We start from requirements, we move to CAD, CAD hands to simulation, moves to manufacture, gets validated, goes to production.

1:10 Sounds great. Doesn’t go that way, right? We see each handoff is more chuck it over the wall. We go from one system to a disparate system to another. Requirements go to CAD, CAD misses something, goes back. Design hands it to simulation, simulation says that doesn’t work, they throw it back, you try again. You guys finally find something that works, and then manufacturing laughs and throws it all the way back, right?

1:40 It becomes an adversarial process to go from concept to reality. Why, right? We sort of hit it. Each of these departments is their own. They’re siloed. It is a throw-it-over-the-wall moment, right? This creates these disjointed, non-continuous approaches. And fundamentally, each side doesn’t understand what the next side needs. As engineers, we’re really good at thinking about part cost. We’re often really bad at thinking about time cost, right?

2:21 All of these start-stops they have costs. 80% of your time is wasted on these loops instead of solving problems. Somewhere in the neighborhood of $370 a year is associated with reworks just cuz, “Hey, I can’t make that.” And this all of this comes together to mean that going from concept to physical, ready-to-go object is somewhere in the 6- to 18-month range. What if instead of these disjointed, siloed, disconnected areas, we had a single space, right?

3:07 That single space is in InfinitForm. It’s one platform where design, simulation, and manufacturability come together at once. How do we do this, right? What does it mean? So, first generation, like the the existing approach to generative design is great at capturing performance intent, right? You say, “I want a bracket. I want a piece that meets these performance goals.” What it doesn’t capture is manufacturing intent, right? But, like we said, that’s one of our big blockers.

3:51 We change that paradigm. Manufacturing intent becomes a fundamental input alongside with that performance intent. Our approach is built on totally proprietary house-built deterministic solvers written from the ground up to run on the GPU. That said, AI is incredibly useful, right? We have a co-pilot deployment with eyes towards MCP, a you know, APIs. But we we try to deploy it where we feel it has the most value.

4:25 Engineers like deterministic solvers. Sure, if you’re exploring a huge space, maybe you like the AI to try to narrow it, but you’re probably going to kick it back to a deterministic solver. Instead, we take the approach of a really fast deterministic solver, but we use that AI for things like importing load sets, matching up how you’ve documented a load set or a constraint set to how our system wants it.

4:50 And then on the backside, it can help you do things like reporting and analysis, you know, kind of secondary analysis. All of this comes together to create a flow that actually looks a lot more like that ideal flow. And from that, we can get to market faster. Not just faster, but with fewer of those adversarial interactions and a lot more collaborative, positive interactions between those departments. So, what is it you know, how does this how does this go?

5:26 We start with inputs, right? This is our intent. Intent is largely three kind of functions, right? So, we’ve got geometry. This is okay, where can my model be? Where are the interfaces? Manufacturing. How do I want to make it? And finally, that performance intent that’s always been there, right? It’s It needs to do this. All of that feeds into our engine. Right? So, I mentioned we’ve got this this AI layer that can help you.

5:57 It’s very optional, but it’s there to help you take that intent in whatever form you have it and translate it into our platform. From there, our solvers take over and optimize for a balanced solution to that problem. Right? It’s not that we’re just re-skinning things. We are fundamentally solving each of those problems in a way that is manufacturing aware. And then for an output, we’re going to actually give you CAD.

6:31 Right? And we’ll see this a little later in more detail. But, we’ll give you real CAD with features. It is going to inherently be manufacturable. Right? Not some after-the-fact check with some bandaids, but a rule-based approach to ensure manufacturability. And that output is going to be performance validated. Right? So, when you hand it If you’re the designer and you hand it to the simulation engineer, it’s no longer a shot in the dark.

6:59 It’s now informed. Right? Same when you hand it to manufacturing. They’re not going to kick it back to you just because you missed a radius. If we look at that, so one of my favorite examples of this is basically a permutation we have where we can set a problem up once and see multiple potential outcomes simultaneously. Right? As a designer, when you go to design a part, historically the status quo is that you have to decide ahead of time how you’re going to make it.

7:37 Right? If I’m going to machine it, if I’m going to print it, if If going to be injection or extrusion, that changes fundamentally how I model that part. Right now, that’s an engineering judgment. Right? You decide ahead of time and you hope you’re right. To be fair, engineering judgment is a little more correct than I think most of us give a give credit. That said, what if it’s a junior engineer?

8:01 What if they don’t have experience in this manufacturing process? Further, as an experienced engineer, you’re aware that you’re guessing. Right? How What if we could move from a guess to an informed decision? That’s what we’re looking at here. So, we’ve set up this one problem and we can see as we loop, we’ll see our design space and then we’ll see some outcomes. Right? Going from left to right, we’re going to have additive, CNC, and extrusion.

8:35 At the bottom, and I probably should have made it a little bigger, we’re going to see performance information on each of these designs. This is going to allow us to make an informed decision. Right? Maybe we’re super super weight conscious of our program. In that case, maybe we pick additive. Right? It is the lightest. Perhaps we want a balanced approach and the 10 g difference here for the machined option is totally within program budgets and not having qualified additive part, not having to pay for that additive part, fits our needs.

9:08 Or maybe we’re doing really high volume and the extruded part makes more sense for even, you know, a little more weight, but a much lower per part cost. That decision that was previously a guess is now an informed decision and it’s one that we wouldn’t necessarily very often have the chance to explore. Right? Very rarely do you have the luxury of redesigning a part after the fact.

9:38 Just to understand if you might want the other version. Right? And what you may notice or may have noticed is that that extruded and machine versions, they looked a lot like B-rep CAD. And that’s because they are. Right? While implicit and mesh are great and interesting and have their own upsides, we fundamentally live in a B-rep world. Right? PLM expects B-rep. Your assemblies expect B-rep. Manufacturing expects B-rep.

10:12 In addition to existing tools not providing you manufacturing intent and constraint at the beginning, you’re stuck with those outputs. Right? If we want to increase the adoption of this sort of approach, we need to make it easy to integrate this approach. And that’s one of the That’s where we see this B-rep output is incredibly incredibly helpful. So, we can we can make steps, but we also are able to integrate directly with NX, SolidWorks, CATIA, and Fusion.

10:49 Creo is on the map. And create actual native CAD with editable feature trees. So, if we look here, we’re going to we’re going to NX. And we’re going to go ahead and pull in a result. And what we’ll see is the creation of sketches and bodies, and we’ll make operations. Each of these sketches is constrained and editable. If you want to change a radius, you can go in and change a radius.

11:40 So, what we’ve seen so far is kind of where we are today. But, what if we extend these ideas? Right? So, I want to give you a somewhat of a tease of where we’re going with this and what we’re working on. What we see here is a drone ribbing solution. So, we’re looking at a roughly meter wingspan drone here. And those ribs are generated as native B-reps in approximately 10 seconds.

12:07 A user can be trained to do this in a matter of hours, not weeks, not months. Down- downstream, these will become optimized ribs for pattern sizing, cross-section, whatever loading you want to put in there. Right? We’re trying to remove the barrier to these optimized geometries and add speed. We’re also working on reverse engineering solutions. Bring us a scan, bring us a drawing, bring us a TDP. We can synthesize that input and create new create the CAD that you’re missing.

12:50 In addition to just creating it, we’re going to be able to validate that that reverse-engineered part at least meets your existing requirements in your new method. Taking that a step further, we look at re-engineering. Right? So, I have a qualified part. I have that reverse-engineered part. I’m going to make it a different way. Why should I just make it the same? Some programs want the same geometry.

13:22 That’s fine. But, wouldn’t it be great if you were able to get the best out of the method you’re using. With the re-engineering approach, you can set a baseline of your existing part and then InfinitForm will explore and find you equivalent performance approaches that are uniquely optimized for the method that you want to explore. In this case, we start with something that’s a casting or a print and we want to say, “Hey, what if I machine it?

13:50 What if I extrude it?” And again, you’re looking at these tradeoffs. You’re looking at tradeoffs or you’re looking at savings and you’re looking at matching geometry to manufacturing. In terms of deployment, we have many options, right? We’re able to deploy completely offline with an on-premise solution that can be run on your local computer, it can be run on your HPC, or it can be run on a private cloud.

14:21 We have people doing each of those. If you do choose to run it, say, on engineers’ computers, because of the GPU native nature of our solver, it is going to give you a much better chance of having your engineers do high-fidelity simulation on a reasonable hardware. For people with less sensitive data, less sensitive applications who are fine with a cloud public cloud deployment, we also offer that, right?

14:48 Completely offload your computing, add collaboration. Ultimately, we’re very flexible, right? And with the on-prem version, we also offer the ability for you to integrate your own LLM, right? So, if you have a model you’ve been working on and it has your company’s intelligence, be that manufacturing guidelines, design standards, what have you, reporting output, any of these things, right? That you’ve already trained your model on, you can just plug it in, maintain all your security, maintain maintain that intelligence, but add it to our platform.

15:28 Right? Fundamentally, we are this layer of the physical intelligence where the AI is not as strong. So, you know, our ask is bring us a problem. We feel that the best the best version of your next design is in our platform. Thank you.