Source description
About the role
About
Rudus is an AI-powered takeoff platform for structural and site concrete. We accelerate concrete takeoffs by 70%+ with AI trained specifically on structural and civil drawings- counting, measuring, and reading sheets so estimators can focus on winning bids.
The Role
You'll own the research behind our drawing-understanding stack: detecting structural and site elements on noisy real-world plan sets, pattern-matching footings and curb runs, interpreting schedules and cross-sheet details, and comparing drawing revisions. Your models ship into production and get evaluated by professional estimators on real bids — feedback loops measured in days, not review cycles.
What You’ll Do
-
Advance detection and autocomplete models for structural elements (footings, slabs, columns, walls) and site elements (curbs, paving, sidewalks, joint layouts)
-
Build document-understanding systems for schedules, specs, and section details, including cross-sheet reasoning
-
Solve geometry problems: auto-scale detection, boundary completion, smart exclusions, revision diffing
-
Design training and eval pipelines on customer drawing data, improving accuracy project over project
What We’re Looking For
-
PhD (or equivalent research experience) in computer vision, ML, or a related field
-
Track record turning research into working systems- strong Python required
-
Experience with document AI, object detection, vectorized/CAD-like data, or multimodal LLMs
-
Excited by messy real-world data over benchmark leaderboards
More at Rudus