Conformal Lattice Design Made Easy: A CAD-Integrated Approach
Interview with Rachel Azulay of Tetmet
In this interview ahead of her presentation at CDFAM, Rachel Azulay introduces Tetmet and its core technology, Adaptive Spatial Lattice Manufacturing (ASLM)—a process that departs from powder-based additive manufacturing by robotically assembling and laser-welding rods into large-scale lattice structures. Azulay explains how Tetmet developed its own Lattice Design Suite (LDS) as a Fusion 360 plug-in to enable conformal, manufacturable lattice generation tailored to ASLM’s unique requirements.
The discussion covers the limitations of existing lattice design software, Tetmet’s CAD-integrated approach, and how LDS manages the data flow from sketches to manufacturable structures.
Azulay discusses the challenges of handling large-scale geometries efficiently, and what Tetmet hopes to both contribute to and learn from the wider computational design community at CDFAM.
Could you start by introducing Tetmet, the manufacturing process, and outlining what you will be presenting at CDFAM?
At Tetmet, our core technology is ASLM—Adaptive Spatial Lattice Manufacturing. Instead of melting powders layer by layer, ASLM assembles rods that are spot-welded by a robot and a laser.
That approach makes it possible to build large-scale, energy-efficient lattice structures that traditional additive processes really struggle with.
Early on, we realised there was no design tool capable of generating the conformal, manufacturable lattices ASLM requires. Because this isn’t our core technology and we needed something up and running quickly, we chose not to build a CAD system from scratch. Instead, we developed the Lattice Design Suite as a plug-in for Fusion 360 using its API. That let us focus entirely on solving the lattice-specific challenges while Fusion handles the heavy lifting of geometry and workflow.
At CDFAM, I’ll be sharing the reasoning behind LDS—particularly why we chose not to make it a push-button automatic tool—and giving a look at some of the directions we’re taking it next.
What are the key limitations you see in existing lattice design software, particularly for conformal lattice generation, and how does Tetmet’s approach address these?
One of the main limitations in existing lattice design software is that it’s tailored to powder-based 3D printing. The typical workflow is: fill a part with a lattice and then print a skin around it. We can’t do that with ASLM—having rods stick out would be mechanically unsound.
Another limitation is that most software isn’t built to optimise for the kinds of geometries ASLM can actually build. We can assemble certain topologies that are impossible or extremely inefficient to print with traditional AM, but the existing tools aren’t designed with those in mind.
On top of that, there are process-specific constraints—for instance, rods can’t be arbitrarily thin—and most automatic conformal lattice generators don’t account for that. Finally, we wanted to make sure engineers’ knowledge and intuition are part of the process. Fully automatic tools often override that, and manual tweaking can be cumbersome.
That’s why we designed LDS to follow the same workflow and feel as volumetric CAD—so it’s parametric, engineer-driven, and manufacturable within ASLM’s unique constraints.
Which CAD platforms does your method integrate with, and how do you manage the interaction between native CAD geometry and lattice generation algorithms?
Right now, we’ve integrated LDS as a plug-in for Autodesk Fusion 360, using their API. Like I said before, as a startup selling a new manufacturing process, we didn’t want to build a CAD system from scratch, so this approach let us get something functional quickly and focus our effort on lattice-specific challenges.
In terms of interaction, we don’t replace the CAD kernel—we work with it.
You start from native sketches and surfaces, and LDS adds lattice-specific operations like placing nodes on the sketch, defining cells, and then the same kinds of commands you’d use for normal volumetric CAD—extrude, revolve, loft create voxels that can then be populated with struts.
That way the lattice stays parametric, editable, and tied to the original design intent, while the heavy lifting of geometry management is handled by the CAD platform itself.
Can you describe the data flow from initial CAD model through lattice generation to a manufacturable design, including any validation or optimization steps?
The data flow in LDS starts with the 2D CAD model: you begin with sketches and surfaces in Fusion 360. From there, LDS lets you place nodes and define cells in 2D space, essentially building the lattice graph on the sketch or surface.
Next comes the 3D stage: the 2D nodes and cells are converted into 3D voxels, which are then populated with struts to form a conformal lattice. At this point, the engineer can manually tweak the lattice as needed. We’re also adding an optimisation layer to guide decisions like rod size, connectivity, or stress alignment without taking control away from the engineer. Finally, the design is exported to FEA software for validation before manufacturing.
So the full flow is:
2D sketches → 2D lattice sketch → 3D voxelised lattice → (in dev) optimisation → FEA validation → manufacturable design.
This approach keeps the lattice parametric, editable, and fully tied to both the engineer’s intent and ASLM process constraints.
How does your approach handle large-scale or complex geometries while maintaining efficiency and manufacturability?
In practice, we’re usually replacing mechanical parts made with traditional processes, so the parts themselves are somewhat limited in geometric complexity. That said, we handle more complex shapes by combining algorithmic and manual approaches.
LDS has algorithms to connect different extrusions and automatically generate lattice connections, and engineers can also manually add, remove, or merge nodes where needed.
One of the big challenges with traditional CAD is the heaviness of B-Rep representations, which makes large lattices slow and unwieldy. That’s why we ‘hack’ the workflow a bit: we create intermediate STL or mesh representations for fast manipulation, and only generate the full B-Rep solids once the lattice is finalised. This lets us handle larger, more complex geometries efficiently while keeping them manufacturable for ASLM.
Concerning ASLM, we’re not limited like in 3D printing to a set buildable volume. We can place our robots on gantries, or our parts on gantries to make theoretically infinite parts.
What do you hope to share with, and learn from, other participants at CDFAM regarding CAD-integrated lattice design?
We’re a deep-tech hardware company, so most of our focus has been on making ASLM work in practice. That means we have a lot to learn from people who are experts in computational design. What we did with LDS was develop something that works best for our specific needs, and I’m genuinely curious to see whether it could be useful to others in the community.
I’m also eager to learn about all the new directions people are exploring—AI-driven design, advanced lattice algorithms, novel optimisation techniques. As a scientist, it’s exciting to see the creativity and innovation in the community, and CDFAM is a perfect opportunity to exchange ideas, learn from other approaches, and hopefully spark collaborations that push the field forward.
As Tetmet advances ASLM and the development of lattice design tools, Rachel Azulay brings a perspective grounded in both manufacturing constraints and computational design opportunities. Her presentation at CDFAM will highlight not only the technical choices behind LDS but also the broader implications for lattice generation in CAD-integrated workflows.
Join Rachel and other leaders in computational design, AI, and advanced manufacturing at CDFAM in New York, October 29–30, to connect, exchange ideas, and see where the field is heading.
