Code-first approach by LEAP 71 targets full-scale manufacturing of rocket engines
Inside the Dubai company’s Noyron model for computational engineering of meganewton-class designs, and why additive manufacturing is the hinge for its ambitions
LEAP 71 is moving quickly to prove that complex hardware can be generated, built, and iterated by software. The Dubai-based computational engineering company is scaling its algorithmically designed rocket engines beyond lab-scale demonstrators, with reference designs that stretch into the meganewton class and component geometries sized for the largest metal powder bed fusion systems now on the market. At the center is Noyron, LEAP 71’s large computational engineering model, which encodes physics, manufacturing rules, and domain knowledge directly into the code that generates parts.
“The hardest challenge remains translating a computational model into real, testable hardware,” said Lin Kayser, co-founder of LEAP 71. “Especially in turbomachinery, where sealing, material fatigue, and transient conditions during startup and shutdown are critical. These are not just design problems — they demand practical testing, iteration, and close collaboration with manufacturing partners.”
In a conversation with VoxelMatters, Kayser described how LEAP 71’s strategy has evolved from the smaller pressure-fed rocket engines, hot-fired in June 2024, to the monolithic copper aerospike engine, which they hot-fired in December that year, an engine type notorious for its complexity, to the much larger and more ambitious engines, which require turbomachinery. The company’s two current reference engines are the Noyron XRA-2E5, a 200-kilonewton aerospike, and the Noyron XRB-2E6, a 2,000-kilonewton bell-nozzle system. Both are generated by the same software “DNA,” as the company characterizes it, allowing architectural variations without redoing the engineering from scratch.
“We have reference designs for anything from tiny engines to the biggest engines flying to space,” Kayser said. “We have validated the manufacturing of it, and that’s a huge step forward because it means we can produce these from computational models and manufacture them at full size. Now it’s a question of testing them and qualifying them.”
From algorithms to test stands
LEAP 71 transcodes engineering knowledge into executable computational models, then uses those models to generate manufacturable geometry along with predicted operating parameters — pressures, temperatures, and flow rates — that can be compared to test data. The process compresses design cycles and shifts engineering effort from one-off manual CAD to refinement of a computational model on a meta level.
“We don’t even own a CAD package,” Kayser said. “It takes about five minutes to execute the algorithm that designs [an engine], and then you have manufacturable files. When we drive back from the test site, we usually already have a new engine in the computer because we plug in the new data, modify some things, and we are ready to go again.”
Early engines focused on demonstrating the feasibility of designing engines without human intervention. They demonstrated the paradigm that you can create a complex, intricate engine with intricate cooling channels and injection systems entirely automatically. After the initial successful test, the aerospike was generated by the same computational DNA, but produced a wholly different geometry. However, the engine adheres to the same fundamental principles. “The first run, you keep short to look at the data and say, OK, the pressure’s OK, is it within nominal parameters?” Kayser said. “Our algorithm gives you geometry and manufacturable files, but it also makes a prediction. The first run is five seconds. Next, you try to achieve a steady state after 10, 20, 30 seconds. Nothing changes after that — you basically are flat. You let it run a few times and see if anything happens that isn’t expected, and then you’re basically good.”
Stress testing follows. “You start throttling down to 25%,” he said. “You have 25% of the fluids, which means you have 25% of the cooling, which is not great. So now you’re stressing the engine. Then you stress it the other way: 100% and then 130%.” Longevity metrics, he added, come only from real missions: “You have to fly them basically.”
Kayser is cautious about timelines but pragmatic about a flight path. Suborbital platforms and “hopper” vehicles — where throttle response matters for vertical takeoff and landing — are near-term venues. “For spacecraft that land, you usually want to have multiple engines because you can turn off some to reduce the thrust,” he said, noting that deep throttling to 5% or 1% is typically unrealistic. “It’s easier to have multiple engines and turn some off to achieve a smooth landing.”
Printing at full scale
If code makes the geometry, additive manufacturing turns it into metal. LEAP 71 has focused on very large powder bed fusion systems to produce monolithic parts that would otherwise require extensive assembly or be impossible to fabricate. The injector head for the XRB-2E6, 3D printed using Nikon SLM Solution’s NXG platform, illustrates the scale: a 600-millimeter-diameter component, matched to a bell nozzle at roughly 1.6 meters in build height.
“We are now routinely working with the largest metal powder bed fusion printers in the world,” Kayser said. “We are printing the bell nozzle, and we are now talking about engines as big as SpaceX’s. We have no way to test it now, but essentially, we have validated the manufacturing of it.” Manufacturing validation, in Kayser’s view, is the critical gate that turns code into credible hardware programs. “That’s a huge step forward because it means we can produce these from computational models and manufacture them at full size,” he said.
The same code-to-metal pipeline is at the base of LEAP 71’s work on hypersonic precoolers, a type of heat exchanger that can cool incoming air, making combined-cycle propulsion architectures viable. LEAP 71 and Farsoon Technologies partnered to design and manufacture a liquid-hydrogen-cooled precooler concept intended for hydrogen-cooled, combined-cycle propulsion systems used in next-generation air- and spacecraft. Combined-cycle engines integrate air-breathing and rocket modes within a single architecture, operating as jet engines within the atmosphere and transitioning to rocket mode at high altitudes and velocities, thereby reducing the amount of liquid oxygen an aircraft needs to carry for the entire ascent.
At hypersonic speeds — Mach 5 and above — air entering the engine heats rapidly due to compression and stagnation effects before combustion. The result is less temperature differential available for expansion, and therefore diminished thrust. LEAP 71’s approach is to remove heat early and with minimal pressure loss. Recent improvements in metal additive manufacturing enable the creation of monolithic, aerodynamically optimized heat exchangers with extremely high internal surface areas. These precoolers rapidly reduce the temperature of incoming air, restoring favorable expansion conditions and making hypersonic combined-cycle propulsion fundamentally feasible.
To demonstrate this capability, LEAP 71 and Farsoon produced a prototype precooler on Farsoon’s FS811M-U-8 metal laser powder bed fusion platform. The machine’s build volume supported a monolithic structure exceeding 1,500 millimeters in height. The part’s size and complexity place it among the largest known high-complexity aerospace components built via metal 3D printing, and it serves as concrete manufacturing proof for hypersonic heat-exchanger architectures. The build utilized SS316L with a coordinated eight-laser strategy to efficiently fabricate the intricate internals and maintain stability across the tall Z-axis.
The design itself was generated by Generation 4 of Noyron, which LEAP 71 describes as a large computational engineering model capable of autonomously creating manufacturable geometry from physics-based logic and high-level intent. In this instance, Noyron produced a “fractal folding” internal structure that maximizes exposure of the hydrogen-cooled surface to the incoming flow within an aerodynamically efficient outer body. LEAP 71 and Farsoon validated the concept by fabricating multiple smaller phenotypic variants — each an expression of the same Noyron “DNA” — before scaling to the full-size component. The company states that Noyron RP, a related model, is a validated system used to produce high-performance space propulsion components.
“The aerospike and bell-nozzle engines we are developing are not separate efforts — they are different phenotypes of the same computational DNA,” said Josefine Lissner, LEAP 71’s managing director and principal architect of Noyron, in related remarks. “This unified approach allows us to explore fundamentally different engine architectures without reinventing the wheel every time.”
A business built on models, not CAD
LEAP 71 describes itself as a computational engineering firm rather than an engine company — at least for now. The business is to build “base models” that capture physics and production rules for a given domain, then specialize those models for customers. Most of its commercial work is not rocketry, according to Kayser, even if engines dominate the public narrative.
“By the way, 90% of this is not rockets,” he said. “The rockets are very visible on our website because that’s all our own stuff. The customer products are not visible on our website, except in rare cases.” Engagements span fusion-reactor components, heat exchangers, and even non-additive projects where computational design still confers advantage. “We’ve done everything from mechanical watches to water waste separation and electrodes.”
The company’s methodology involves translating expert knowledge into code, generating geometry and predictions, selecting the most suitable manufacturing process, and closing the loop with test data. One provocation in Kayser’s message is that the additive manufacturing industry’s biggest bottleneck is not machines or materials but design methods.
“Nobody knows how to design for AM, and people have been trying to use the wrong tools to do that,” he said. The remedy, in LEAP 71’s telling, is computational engineering that treats geometry as a byproduct of models rather than an artifact to be drawn and redrawn in a traditional CAD system.
That approach extends to how the company engages with the broader AM supply chain. Kayser is vocal about the need for AM to leave its own echo chamber and integrate with mainstream manufacturing disciplines — from heat treatment to welding and fasteners — because that is where complete systems are realized. “You can print something, but it needs screws, bolts, heat treatment; somebody needs to weld something to it,” he said. “Why is this separate? The biggest shame of our industry is that they are in this kind of echo chamber, always looking at only their competitors and completely forgetting that there’s a much bigger market, which is all of building stuff.”
LEAP 71 has also released key pieces of its stack as open source, notably its PicoGK geometry kernel, to encourage adoption and scrutiny. The move aligns with the company’s argument that computational engineering is both a discipline and a product. The more engineers work from shared, inspectable code, the more reproducible and auditable the results can become.
Why the UAE, why now
LEAP 71’s decision to base itself in Dubai is part practical and part philosophical. Practically, the company’s customers are global, and the UAE is courting advanced manufacturing and space companies with an emphasis on execution and infrastructure. Philosophically, Kayser said the local culture of “building things” matters.
“We are surrounded by people who just want to build amazing stuff for the future,” he said. “It is so interesting to see this because you come here and everybody talks about, ‘We had this idea, and then we built it.’ You are amazed because you’d forgotten that humans can make a decision and build something.”
He argues that the UAE offers a clearer and more navigable export framework for propulsion technology than jurisdictions with sweeping restrictions, while still applying controls. “If I want to export a rocket engine to Europe, I need an export permit,” he said. “It’s a clear process; you pay your fee and that’s it.” By contrast, he notes, most U.S.-sourced propulsion hardware is effectively inaccessible to non-U.S. programs due to stringent export regimes.
Accessibility is drawing teams that want to build orbital vehicles but lack the necessary engines — or the ability to source them. Kayser pointed to a recent wave of launcher startups relocating operations to the UAE, including a company led by veterans of legacy Soviet programs. The dynamic suggests a potential reconfiguration of the space supply chain, with propulsion specialists serving multiple launch-vehicle companies rather than each launcher firm building engines in-house.
“Everybody’s trying to copy SpaceX,” he said. “SpaceX built everything vertically because there was no supply chain. But if you think about airplanes or other mature fields, there’s Boeing who builds airframes, and there are engine makers, and then there are airlines that operate the planes. The engine is so complicated and such a special, separate thing that it is almost logical to have it as a separate business.”
That separation could also involve LEAP 71 more directly if the company spins out or seeds a dedicated engine manufacturer. “What’s next for us is very likely to build another company that actually builds rocket engines — like physical ones,” Kayser said. “It’s not clear if this is a company we would run as CEO or just provide technology and own a part because it’s a different business. For us, this is a likely next step.”
While the company is self-funded — Kayser notes that a single test campaign costs the equivalent of a mid-range luxury car — it prioritizes projects where progress is a function of code, materials, and machine capability, rather than marketing cycles. LEAP 71’s calendar is anchored to manufacturing and test windows more than conference stages, even if trade shows remain useful for situating additive manufacturing within larger industrial narratives.
That pragmatic cadence is visible in how the team frames milestones. Printing a 1.6-meter-class bell nozzle on a large-format metal system is not a trophy so much as a validation step. Manufacturing the 600-millimeter injector head is a precondition for integrated hot-fire assemblies. The near-term testing focus is on simpler cycles, such as gas generators, to build a robust foundation before scaling to full-flow staged combustion. The model evolves with each run, and the hardware follows, sometimes literally overnight.
“Parameters create different things,” Kayser said, holding up a very large part during the interview. “This one is two Meganewtons. And that’s only the upper part.” The throughline is that the parts are not hand-drawn variants but expressions of a living codebase that can be steered to different outcomes under explicit constraints, including manufacturability.
LEAP 71’s leaders are realistic about the road from showpiece to service life. “They will not work perfectly in the first go,” Kayser said. “So far, every engine we tested worked, which is unusual — but we’ll see.” The next critical proof is sustained, repeatable performance in relevant environments, followed by throttling profiles, and then integration. The model will continue to generate candidates; the printers will continue to make them; and the test stands will continue to instruct the code on what to learn next.
