“We think engineering is in a crisis,” Lin Kayser says of he and Josefine Lissner, sat to his left.
So much has changed since he last made time to speak with TCT on the record, but the mission hasn’t, and nor have the emphatic observations. Kayser first provided his frank assessments of the engineering world to TCT in February 2021, where he noted the need to ‘dramatically accelerate innovation.’ He then had the conversation with EOS founder Hans Langer on the very first TCT Innovators on Innovators podcast. Before all that, he would tell anyone who would listen, be it readers of his blog, attendees of his presentations, or investors in his company.
Lissner hasn’t been at it for quite as long, but she is no less adamant. The pair met at Hyperganic, the computational engineering company Kayser founded with CTO Michael Gallo in 2013, and departed earlier this year for much the same reason. It was Lissner who left first. In January, she decided to set up her own computational engineering services firm, Leap 71. The idea at that stage, with Kayser still at the helm of Hyperganic, was that Leap 71 would use platforms, such as those Hyperganic were bringing to market, to develop the algorithms that would develop the parts.
That is no longer the case. Kayser is now considered a co-founder of Leap 71 having left his role as Hyperganic CEO in March. His departure came after disagreements between C-level execs and investors around how to take the company forward. Kayser and Gallo were committed to pursuing their admittedly ‘crazy vision’ of revolutionising engineering, while the investors saw an opportunity to generate immediate revenue through Hyperganic’s ‘fantastic voxel kernel for lattices.’ The founders lost the argument, and though they had the opportunity to take a step back into honorary roles at the company, receiving credit for developing the tech and founding the company, they would no longer be in control.
“Sorry, no. That’s not what I want to do for the next decade of my life,” Kayser remembers thinking. “The other thing is, you could see that it started to work, right? You saw the aerospike [engine] that Josie did, and we announced a collaboration to build air conditioning units. Stuff that has an impact. I wouldn’t relent, and now I’m out, and Josie’s out, and we said, ‘What shall we do?’”
Moving forward as co-founders of Leap 71 was the answer to that question. Lissner had already made some headway, launching the brand on socials, developing a following of more than 1,000 people on LinkedIn alone and, more importantly, writing lines upon lines of code.
Get your FREE print subscription to TCT Magazine.
Exhibit at the UK's definitive and most influential 3D printing and additive manufacturing event, TCT 3Sixty.
What Leap 71 is developing is a stack of kernels, with the foundational and open-source PicoGK as its bottom layer. As you go up the pyramid of kernels, there are more application-specific modules – like RP/CEM – but the bottom layer is the ‘generalised stuff that no engineer should spend time on again.’ The company leans on Open VDB for its voxel engine and an API that has only 20 functions. It ‘supports the same vision’ as Hyperganic’s IP but has been developed with floating-point math rather than the fixed-point math that was used to build Hyperganic’s Core platform. The kernel also only has around 100 lines of code, with Leap 71 taking inspiration from ARM’s Reduced Instruction Set Computing (RISC) microprocessors that are simple and streamlined but fast and energy efficient.
“This is kind of what we’re trying to replicate because when you look at most software out there, they have 1,000 functions that do exactly one thing,” Lissner says. “They are very specialised, so they are never always exactly what you want, where we have very foundational functions, nothing fancy, but they’re very fast and very reliable. That’s the idea. And then someone can come along and build all the sophistication on top of it, which is my job.”
Leap 71
As Lissner sets about building out the technology stack, decisions will be made as to which modules are made open source, while through its own application of the kernels, the company hopes to see electric motor coils, propulsion components, and perhaps even rocket engines developed if the recently announced Exploration Company collaboration progresses well.
Exploration Company rocket engines would be designed using RP/CEM, with Leap 71 inputting information like the desired thrust level, available chamber pressure, and the maximum expansion ratio, as well as manufacturing constraints like wall thickness and overhang angles, to generate an engine design and a print file. Because of these inputs, it is unlikely that the algorithm will ever output parts with overhangs or wall thicknesses or support structures that are not printable.
RP/CEM runs on PicoGK, but sitting many layers up the pyramid, it does not interact with the foundational kernel. When working on projects like engines and propulsion systems, it has taken Lissner a matter of hours, potentially a full working day, to generate a first iteration. But the second and third iterations should, in theory, take only minutes thereafter. The actual generation of the application once the code is written might only take 15 minutes, but that can depend on the resolution. For first iterations, Lissner works with a low resolution, only rendering in the higher resolution when she’s ready to print. What’s more, the intelligence of the user will be baked into the algorithm, so no danger of tribal knowledge being lost or forgotten, and there’s also the potential for lines of code to be repurposed for other projects or tweaked to enhance performance.
“It preserves the thought process, rather than just the visual result,” Lissner explains.
Kayser adds: “The thought process, not the blueprint. We have all the blueprints from the Apollo programme [for example]. Nobody knows why they did it like that. That’s a huge problem in engineering right now. It’s in the brains of a few people and nobody understands why they made that choice. They can [only] guess.”
Leap 71
RP/CEM will be kept proprietary by Leap 71, with companies like The Exploration Company – and four or five others so far – able to benefit from it through the company’s service offering. PicoGK, the platform from which RP/CEM was built, however, will be open to all. Lissner describes PicoGK as having ‘very foundational, very basic geometric stuff’, with a shape kernel sitting on top of it – also to be made open source. The shape kernel allows the user to build basic shapes like cylinders and modulated cubes and sits between where a user can control individual voxels and where the rocket nozzle, for example, would be created.
The point of making PicoGK open source is Leap 71 feels the layers that sit on top of the kernel aren’t amazingly complex essays of code, and so it should be feasible for others to build out computational engineering capability for their own needs too. Lissner and Kayser are cognisant of the challenges involved in altering the way engineers work and the way parts are developed, but they are not perturbed.
With the right software tools, the pair are convinced that a significant dent can be made.
“People don’t realise how much a good software person scales compared to, for example, in the hardware space,” Lissner says. “In the hardware space, if you have a good hardware person, you might be two times or three times as productive as [an average person]. So, you still end up with a lot of workforce. Software companies can scale completely differently if you find a bunch of the right people because they can do stuff that other people cannot do at all.”
“It’s just incredible how slow engineering is,” Kayser adds, “and it’s kind of sad because these are very smart engineers. If you give superpowers to engineering, the world will change.”
They agree that engineers are being held back by the technologies available to them, and even suggest that an open-source platform for computation engineering could help retain skills within the industry owing to a new opportunity to get creative. The pair are placing great stock in the capability of engineers and also their willingness to push the boundaries with new technology; perhaps even their frustration with the tools they have at their disposal.
Leap 71
“This will change what we perceive as design for additive manufacturing, because when you look in the industry right now, it’s gyroids, it’s topology optimisation. Okay,” Lissner says, “topology optimisation is fine, but I never use gyroid as the centrepiece for any of my designs. It’s the last 5% if I’m like, ‘Maybe I should lightweight it slightly,’ then I use a gyroid. It’s 0.1% of what I see as the value of additively manufactured design. This drives me crazy because heat exchangers are gyroids, lightweight parts are gyroids, heat sinks, everything is a gyroid. This is just what the software can do.”
“Because of this prevalence of implicit function, because it became synonymous for design for additive manufacturing, everybody’s trying to build everything from implicit functions, which means repetitive surfaces,” Kayser says. “Implicits are very, very powerful, you can use them for a lot of things, but they are also very limited.”
“As the name says, you implicitly describe the shape, you don’t explicitly describe the shape, and it just absolutely makes your mind explode when it is something remotely complex,” Lissner adds. “So, there is a big limitation if you want to create shapes.”
Leap 71 is thus coming to market with its technology stack under the staunch belief that a computational approach to design and engineering is what can unlock the potential of additive manufacturing. CAD tools might enable engineers to optimise topology and integrate lattices, but there is so much more that existing CAD tools can’t facilitate that there is cause for change.
While Leap 71’s focus is on transforming the way parts and systems are designed and engineered, that there also needs to be improvements in additive manufacturing hardware hasn’t escaped their notice. If, as Lissner and Kayser hope, their vision of a computation-engineered future comes to fruition, then hardware manufacturers are going to have to keep pace. And if, as the industry expects, additive manufacturing is to become a bona fide manufacturing method, then they also need to make more headway in terms of automating the rest of the workflow.
“People have talked about industrialisation for a long time, but there seems to be no active driver,” Kayser says. “Additive manufacturing is a boutique way of manufacturing individual objects. There’s always some guy running around with a vacuum cleaner, and there’s people with forklifts, and emptying build volumes. I would argue it’s a science experiment right now. It’s not a manufacturing process.”
Since the constraints of the manufacturing process are baked into the coding, Leap 71 says the modules will be able to adapt as improvements are made to the hardware on the market to output more capable geometries. This, the pair suggest, is a ‘recipe for innovation’ – the thing Kayser noted needed dramatically accelerating two years ago.
Leap 71
Though their vision for designers and engineers to embrace algorithmic engineering might be wild, Lissner and Kayser are sure to regularly caveat that this kind of transformation is a huge undertaking for a community. In any case, they don’t expect to convince every engineer; a small minority – or a ‘few young engineers who are bored to hell’ – might be enough to kickstart the change they want to see.
This, Kayser says, is why PicoGK needs to be open source. Taking inspiration from hobbyist open-source movements, Leap 71 suggests there is a need for the same level of sharing and collaboration in the professional space.
“I think this is probably the first time in the last seven, eight years, that a paradigm comes to the market that makes it broadly accessible to everyone to design all the spaces and sophisticated things that we've shown over the past half year,” Kayser suggests. “And I mean, we are not the only ones who can think of interesting things to build. There are millions of people out there who have access to this now, you don't need a fast computer for it. You don't need a special graphics card for it. You don't need an expensive software package for it. You can just download it and get started.”
It's apparently as easy as that. Everything that can be seen on the Leap 71 website currently has been designed and manufactured in the last six months. Lissner has taken the lead on the engineering of those new products, utilising the computational engineering algorithms she has coded this year on the cheapest Macbook Air model on the market.
Their computational engineering vision might seem far-fetched, but the generation of complex parts, products and systems needn’t be so difficult.
“PicoGK was created in three months. It’s 100 lines of code. It’s not like we are geniuses, you just need a very thin layer,” Kayser summarises before Lissner adds: “You just need the right 100 lines of code, not any 100 lines of code.”
