2025 · 6 months · Packhunt
Computational Designer, Technical Lead
Parametric stair configurator built in Grasshopper that generates multiple stair types with clean geometry and applies customer-defined rules. The system uses native components, solid data tree management and real-time feedback to make stair design faster and more reliable.
Stair compliance gets checked at the end, on drawings, by hand. By then a violation is expensive: it means coordination between architects, engineers and reviewers to fix something the geometry should never have produced in the first place.
Four stair types with different rules and tangled interdependencies, plus customer-defined constraints that had to be enforced without leaving the designer unable to see why a configuration had been rejected.
A parametric configurator that routes all four typologies through common filtering logic at the head of the definition, then applies rule-based validation with real-time warnings in the viewport. Native components and disciplined data trees throughout, and documented clusters with typed inputs and stated stress cases so the file could be handed over without a meeting.
Computational designer and technical lead. I owned the architecture, the rules engine, the warning behaviour, the documentation, and the alignment between architects, engineers and project managers.
Compliance stopped being a review step. A rule violation now shows up in the viewport while the stair is being configured, instead of in a drawing review weeks later, which removes a whole class of back-and-forth between architects, engineers and reviewers. The weekly check-ins kept the rule set and the technical work aligned as it grew.
In architecture, automation isn't really about speed. It's about catching mistakes before they reach drawings, by baking expert knowledge into the tool itself. Agreements in documents and meetings age; a validation rule does not. A configurator that only outputs compliant stairs takes a whole class of coordination problems off the table.
The early weeks mattered more than I expected. Time spent mapping requirements, product specs and the awkward edge cases nobody wanted to think about paid back many times over, and splitting four stair types into modular pieces meant we hit the unknowns one at a time instead of all at once.