Autodesk University 2026
Your Firm’s DNA in a Revit Skill: AI That Learns and Replicates Your BIM Standards
Cheli Hershcovich (BIM Developer, SWAPP) with Kari Schutte (Senior BIM Administrator, Stantec)
Submitted for Autodesk University 2026 — not yet scheduled.
Dimension DNA: automated exterior dimensions in a firm’s style
An AI system reads a client’s Revit model, extracts the dimension logic encoded in it, and populates a JSON file that a pre-built parametric skill reads directly. The result: any team member can document to firm standards from day one — replacing 2–3 weeks of mentoring with a single reviewed step.
What you’ll be able to do
- Analyze how AI extracts exterior-dimension logic from an existing Revit model (family type, string count, spacing, and reference targets) to identify which JSON fields need human verification before the skill runs.
- Evaluate the JSON output of a model-analysis pass and assess whether the extraction is accurate before running the skill.
- Identify which documentation standards in a typical project model are reliably extractable today versus which need better model quality or manual input.
- Apply a pre-built parametric skill loaded with a client JSON file to generate exterior dimension strings on a new model — reducing a 2–3 week knowledge-transfer process to a single reviewed step.
Session outline
-
Opening · 8 min
The Standards Problem
Live, no setup. A Revit model from a real residential project goes in. Seconds later, a JSON file comes out: dimension family name, string count, spacing values, reference targets, all readable on screen. We inspect it, load it into a pre-built parametric skill, and run it on a different model. The skill places exterior dimension strings in that firm’s exact style.
-
Section 1 · 12 min
What the Model Already Knows
Why documentation standards are rarely written down, and why the model is the most reliable record of how a firm actually works. The difference between surface-level model data (element counts, parameter values) and structural documentation logic (dimension string composition, annotation family selection rules, per-view annotation sets). Where template-based approaches fall short.
-
Section 2 · 22 min
Live Workflow: From Model to JSON to Drawing
A step-by-step walkthrough on a live residential model: (1) model upload & analysis, (2) JSON review on screen: family type, string count, spacing values, reference targets, each field explained, (3) skill run on a new model, (4) output review & gap check. At each step: what the system was given, what it decided, and where human review mattered.
-
Section 3 · 10 min
Setting the Quality Bar for Extraction
Three model conditions that produce unreliable extractions: inconsistent family usage across levels, overridden instances that contradict type-level logic, and models produced by multiple firms with conflicting conventions. How to prepare a source model for a clean extraction pass, and the difference between a skill that captures a firm’s actual standards and one that captures their worst habits.
-
Closing · 8 min
What’s Extractable Today
A one-page decision map: which dimension-logic patterns are reliably extractable today, which require better model quality first, and which annotation types are the natural next candidates after exterior dimensions. Q&A.
Speakers
Cheli Hershcovich
BIM Developer, SWAPP AI
Cheli Hershcovich designed and built the extraction pipeline and parametric skill architecture demonstrated in this session. She holds an MArch (Technion), spent 8 years in practice at firms in New York and Israel, and builds Python-based automation tools validated on real client models across educational, clinical, data-center, and student-housing project types.
Kari Schutte
Senior BIM Administrator, Stantec
Kari Schutte ran this workflow in production, managing Revit standards at the firm. An Autodesk Certified Professional with a B.Arch (University of Washington) and 30+ years of AEC experience, she provides direct practitioner validation — the client who ran the technology on live projects.