On the floor
We scope, deploy and debug in fab shops. Assumptions die quickly there, which is the point.
Weldeon is a small, technical team building autonomy for a process that punishes hand-waving. If you want your work measured in first-pass yield rather than engagement, read on.
Most software fails softly. A retry, a rollback, an apology in a status page. Ours writes parameters to a machine depositing molten metal into a joint that will hold a bridge up for fifty years.
That changes how we build. Boundaries are enforced below the intelligence, not above it. Every action is recorded before the next one begins. Confidence is a first-class value, and 'we are not sure' is a supported outcome that stops a cell.
It also changes who enjoys it. People who like constraints, physical reality and being told plainly by a welding foreman that they are wrong tend to thrive here.
Six practices we actually hold to.
We scope, deploy and debug in fab shops. Assumptions die quickly there, which is the point.
An engineer owns a scope end to end — sensing, model, adapter and the number it moves.
Welding and quality engineers sit inside product decisions, not downstream of them.
Design decisions are written down and argued in text before they are built.
We label what is unknown internally with the same discipline we use externally.
Meeting load is deliberately low. Physics does not respond well to context switching.
All roles are hybrid with regular travel to customer fab shops. [PLACEHOLDER] — role detail and locations are confirmed in the first conversation.
Sensor fusion across laser tracking, through-arc sensing, thermal and vision to produce a metric joint estimate with honest uncertainty.
Real-time parameter and motion control on the factory edge, inside a hard qualified envelope, at 20 ms.
Own procedures, envelopes and code models. Translate D1.1, BPVC, ISO 3834 and EN 1090 into machine-enforceable constraints.
Physics-informed models for weld sequence, distortion and residual stress that run fast enough to plan a shift.
Radiographic and ultrasonic interpretation with calibrated confidence and an inspector genuinely in the loop.
Own integration weeks: controllers, sensors, MES, roles and the baseline that makes a scorecard mean something.
Sell a paid pilot to a fabrication operations director on a scorecard, not a demo.
Four conversations, about two weeks, no take-home marathon.
Depth
few people, whole problems, long attention spans
Floor
regular time in customer fab shops for every engineer
Written
design decisions argued in text before code
Honest
unknowns labelled internally and externally
Team size, funding stage and specific locations are shared in the first conversation rather than published.
“Physics does not care how confident your model sounded.”
Concrete, because vague career pages help nobody.
Ship something small to a real cell, and spend at least one week in a fab shop watching people weld.
Own a scope end to end — the sensing, the model, the adapter and the metric it is judged on.
Lead an integration week at a customer site, including the parts that go wrong at 6am.
Own a scorecard line publicly, in front of a customer, including the weeks it does not move.
[PLACEHOLDER] — final benefit detail confirmed at offer stage.
Meaningful ownership with the calculation explained rather than implied.
Access to cells, sensors and edge appliances. You cannot build this on a laptop alone.
Fab-shop visits are funded and expected, not rationed.
Including welding certifications for engineers who want to actually understand the process.
No cover letter. Tell us about a hard technical problem you owned end to end, and which of our roles it points at.
For the welding-engineer and NDT roles, yes. For perception, controls and simulation, no — but you need genuine enthusiasm for learning the process, because you will be on a fab-shop floor within a month.
Hybrid with regular travel. The product is built where the metal is; teams that stay remote from the floor build the wrong thing very efficiently.
Write anyway. Small teams hire the person before the requisition more often than the reverse.
Book a pilot on one cell or line. We agree the scorecard before we start — first-pass yield, rework hours, defect escapes, takt — and we report against it in the open.