In our Q1 2024 quality audit, I reviewed a build file that, honestly, almost slipped through. It was an acetabular reamer handle for a medical device company. The kind of part that looks simple at first glance: a long, thin shaft, a threaded interface at one end, a cutting head attachment at the other. They had been machining it for years—all solid titanium, conventional reamer CNC work—and it was heavy enough that surgeons complained after a few hours in the OR.
I’m the person at Velo3D who reviews build files before they reach a Sapphire printer. I don’t have a glamorous title. But essentially, I’m the last line of defense between a CAD model and very expensive scrap. Over 4 years and something like 250 unique parts a year, I’ve learned that most quality issues don’t announce themselves when you open the model. They hide in the assumptions.
This particular part, though, was not a geometry problem I could catch by eye. It was a communication mismatch that took a simulation to uncover.
From CNC reamer to a lightweight design
The client—a mid-size orthopedic device manufacturer—came to us through a design-for-additive-manufacturing consultancy. Their goal was straightforward: keep the same surgical interface dimensions, cut the weight, and add a lattice structure in the grip area for ergonomics.
Because their team was coming from a CNC world, the CAD file carried a bunch of unstated assumptions. Uniform wall thicknesses. Full fillets. No internal features. Our application engineers did a first pass and told them, “This will print, but you’re not getting the real benefit of AM with this design.”
Then the conversation got complicated.
We both said “no supports.” We were not talking about the same thing.
Here’s where I have to be fair to the client: our own material on the Velo3D official website talks a lot about printing complex geometries without supports. That’s a real capability. The Sapphire platform handles unsupported overhangs, internal channels, and thin walls in ways that conventional powder-bed machines struggle with. And the Velo3D–SpaceX contract 2025 cycle includes parts that basically depend on that capability.
But “the machine can print this geometry without support” is not the same as “your part, in this orientation, needs zero engineering thought about supports.”
I said: “Your lattice area will print fine unsupported.”
They heard: “We never have to think about supports again.”
We discovered the gap when they submitted a revision with an extended cutting head that overhung the build plate at a steep angle. The simulation flagged it immediately. So did our pre-print verification tool. If the operator had loaded that file and hit “build,” the part would have printed with a visible distortion in the cutting face region. The thermal model predicted just over a millimeter of deviation in a feature with a 0.1 mm tolerance. That was not a cosmetic problem. That was a scrapped build, a wasted titanium lot, and a clinical trial schedule blown by two weeks.
When the client asked about injection molding
Somewhere in the middle of the project, the client’s purchasing lead asked: “Why don’t we just injection mold this handle? Seems like it would be cheaper.”
Honestly, it’s a fair question. In fact, “how does injection molding works” is one of the searches that lands people on our website all the time. So, for anyone still wondering: injection molding uses a hardened steel mold cavity, into which molten material is forced under high pressure. That cavity is custom made for each part, and the tooling is expensive enough that you need very high volumes to make the math work. For a surgical handle with annual demand in the low hundreds, the tooling alone kills the business case. Add the internal lattice geometry, and injection molding was never realistic.
The actual decision was between CNC—which gave them a heavy, solid handle—and metal AM, which could give them a lighter one with internal structure. I explained the distinction and we moved on.
Eighteen hours of verification vs. eleven days of correction
When the simulation flagged the overhang, the project lead asked if we could just print it and see what happens. I get why he asked. The clinical trial window was closing, and every day mattered. There’s a temptation in every engineering organization to let a marginal result slide because somebody is waiting on the other end.
That is exactly where the prevention-over-cure philosophy has to win.
We asked for 24 hours to adjust the orientation and re-run the simulation. The actual work took about 18 hours of engineering time across two people. The compute cost was negligible. The output was a revised orientation that cut the predicted distortion from roughly 1.1 mm to 0.09 mm—comfortably inside tolerance.
I still remember watching that build come off the Sapphire printer and go through inspection perfectly. We shipped on time. The client never experienced the failure they were almost walking into. And honestly, that’s the part of my job I like: nobody can prove what didn’t happen, but we all know what it would have cost.
If we had printed the broken file, the build would have run for several days, then failed metallurgical testing, then required a second run. The rework cost would have been in the tens of thousands of dollars, not including a two-week schedule hit for the client’s trial. The fix was 18 hours of engineering. Prevention over cure, every time.
The lesson I keep re-learning
It took me 3 years and roughly 150 build-file reviews to realize that the highest risk in metal AM is not the printing—it’s the handoff from design intent to build reality. Every department involved had good intentions. Nobody made a reckless decision deliberately. The problem was shared vocabulary, and the safety net underneath it all was the verification step.
In 2022, one of those handoff failures cost us a $22,000 rework and a delayed launch. That event produced my 12-point inspection checklist. Since then, I’d estimate the checklist has saved us at least three builds from the same fate. Not a bad return on a document.
So, if anyone is working with an AM supplier, here are three things I keep telling myself:
- Never skip the build simulation on parts with thin walls, unsupported overhangs, or internal channels. It is the cheapest inspection you will ever run.
- If a supplier says your design is printable, ask to see the build strategy and simulated distortion values. If they can’t produce those, that’s a red flag.
- Write your own verification checklist. It doesn’t have to be 12 points, but make it mandatory.
5 minutes of verification beats 5 days of correction.
That sentence is the philosophy in one line. We apply it to every part we review, from a 20-part medical order to the Velo3D–SpaceX contract 2025 reviews running through our quality department this year. It’s all prevention, all the time.
And if you’re evaluating metal AM for your own parts, the Velo3D official website is a useful starting point—you’ll find printer specifications, material data, and case studies there. Just don’t make the same mistake our client almost made: don’t confuse a datasheet with a buildability review. The datasheet tells you what a material can do in theory. A proper review tells you what your specific part will do in practice. That gap is where quality problems are born.