Algorithm Engineer

Robotic Imaging

Turning messy physical reality into Impossible Intelligence for the world's largest retail portfolios. Business HQ · Miami · Eng HQ · Tel Aviv · In-person · Profitable & self-funded.

The bridge between modeling and intelligence.

Robotic Imaging turns scans of retail sites into precise BIM. That modeling business does multi-million dollars in revenue. On top of it, our intelligence team is building Claude for enterprise retail: the predictive layer over the physical record. You are the bridge. You own the underlying data, the pipeline that turns every point cloud and Revit model into structured, queryable truth, and you automate the modeling that produces it.

We hold 5,000+ historical projects pairing raw point clouds with finished Revit models — one of the rarest commercial datasets anywhere. Your job is to make it both a faster modeling engine and the foundation our intelligence runs on.

The challenge

• Inputs: registered point clouds and 5,000+ finished Revit models from retail scans — arbitrary coordinates, real-world noise.
• Goal: automate modeling (point cloud → native .rvt, 2D CAD, PDF floor plans) and turn the output into structured data the intelligence layer can query.
• The catch: the post-processing layer — turning segmented points into clean, orthogonal, design-intent geometry. Every off-the-shelf tool we tested failed here.

What you'll do

• Build the point-cloud-to-Revit pipeline end to end: segmentation, geometry extraction, native Revit generation.
• Own the geometry layer: best-fit straight and orthogonal walls within tolerance, family matching, per-client coordinate alignment.
• Own the data contract: turn every model into the structured, canonical record our intelligence team builds on.
• Ship per-element confidence scoring so low-confidence regions get flagged, never shipped blind.
• Hit the bar: 95% of elements within 1 inch tolerance — output a modeler refines instead of rebuilds.

Must have

• Strong ML + geometry: point-cloud segmentation, 3D deep learning, computational geometry.
• Computational-geometry judgment: best-fit straight/orthogonal geometry, tolerance reasoning.
• Strong software + data modeling — you think in data models, not just geometry, because downstream systems have to trust your output.
• Quality obsession — you know a sloppy starter file is worse than a blank one.
• Zero-to-one builder who takes a whiteboard idea to production in record time.

Nice to have (or fast to learn)

• Autodesk / Revit ecosystem and native .rvt generation. If you've done this, we weight it heavily — it's the hardest piece to learn on the job.
• Per-client coordinate alignment.
• Retail or built-environment domain experience.

Do not apply if you need a ticket queue or detailed specs handed to you. You'll navigate ambiguity and build.