UNSTRUCTURED CAD
Industrial STEP models rarely arrive with every engineering interface cleanly identified.
Autonomous Engineering
RouteForge AI is building an agentic engineering platform that transforms industrial CAD and requirements into validated, auditable designs — starting with autonomous piping.
Agentic decisionsDeterministic engineeringHuman authority
skid_A12.step — ROUTEFORGE WORKSPACE
RUNNING
Animated illustration of the RouteForge workflow: a CAD model is ingested, connection interfaces are detected, the engineer confirms source and target, a collision-aware pipe route is generated, and clearance and collision validation pass.
STATE 01/06 — UPLOAD CAD
The Problem
CAD geometry, process intent, equipment interfaces, engineering requirements, clearances, specifications, and simulation often live across disconnected tools and human knowledge. Traditional automation frequently assumes these inputs have already been structured.
RouteForge is designed to reason over that gap — while keeping engineering truth grounded in deterministic tools.
Industrial STEP models rarely arrive with every engineering interface cleanly identified.
Engineers spend significant time interpreting geometry, selecting interfaces, routing, checking, and iterating.
Geometry, requirements, routing, validation, and simulation often exist as separate workflows.
Automation becomes dangerous when missing information is silently guessed.
The Approach
RouteForge maintains a continuously evolving engineering evidence state. Every capability the agent runs adds measured facts to that state — and every downstream decision is made against it, not against assumptions.
Evidence State — live during workflow
How RouteForge Works
The agent sequences these capabilities autonomously, re-entering earlier steps whenever new evidence changes the picture.
Inspect components, topology, surfaces, dimensions, bounds, and representation quality.
Identify defensible physical connection candidates using analytic and derived geometry.
Combine CAD evidence with engineering requirements such as pipe size and minimum clearance.
The engineering agent determines which deterministic capability should execute next.
Produce a collision-aware route between confirmed engineering interfaces.
Convert the route into parametric engineering geometry.
Check collisions, clearances, connectivity, geometry integrity, and requested simulation conditions.
Expose assumptions, evidence, provenance, unresolved semantics, validation results, and the final design.
Architecture
The agent never invents dimensions, geometry, clearances, pipe sizes, or simulation results. It requests capabilities across an evidence contract — and deterministic engineering software returns what is measurably true.
Engineering Agent
Decides what should happen next.
Output: agent action
Deterministic Engineering
Computes what is physically true.
Output: measured evidence
SUFFICIENT EVIDENCE
→ Proceed autonomously.
AMBIGUOUS EVIDENCE
→ Investigate further.
UNPROVABLE SEMANTICS
→ Ask the engineer a bounded question.
Autonomous where evidence is sufficient. Human where engineering judgment matters.
CAD Intelligence
RouteForge is being built for engineering geometry beyond perfect demo models — the STEP files that actually leave vendor tools and translation pipelines.
Clean cylinders, planes, and rings with explicit analytic structure.
Seam-split geometry, partial arcs, and mixed topology.
Converted surfaces, fragmented interfaces, incomplete analytic structure.
RouteForge Geometry Intelligence
When geometry cannot defensibly support an engineering conclusion, RouteForge reports the limitation instead of fabricating an answer. No system understands every CAD file — the difference is whether it admits it.
Human in the Loop
Geometry can identify possible physical interfaces. But geometry alone cannot always determine process semantics — such as “Which nozzle is the source?”
Instead of silently choosing, RouteForge pauses with a bounded engineering question. The engineer answers from the evidence already gathered, and the workflow resumes exactly where it stopped.
Engineer confirmation required
RouteForge will not guess process semantics.
Routing
Once source and target are confirmed, deterministic routing generates an orthogonal pipe path that respects keep-out volumes and the required minimum clearance.
PLAN VIEW — ROUTE GENERATION
DN100
Route Result
Illustrative values from a single prototype run — shown as product UI, not as universal performance claims.
Validation
A design should be checked before it is trusted.
A generated design is not automatically a validated design.
Detect intersections with equipment and keep-out geometry.
Verify minimum separation from surrounding geometry.
Confirm route attachment to selected interfaces.
Validate generated route topology and CAD integrity.
Track requirements against measured evidence.
Orchestrate deterministic analysis where requested and supported.
RouteForge separates workflow completion from engineering validity.
Passing checks builds a case — it does not replace judgment. Final professional approval remains with the responsible engineer.
Auditability
Auditability is a first-class architecture feature in RouteForge, not an afterthought. Each action records what ran, why it ran, what evidence it produced, and who confirmed what.
The result is a deterministic action trace: a reviewable record connecting requirements, geometry evidence, agent decisions, human confirmations, and validation results — for every design the system produces.
AUDIT TRAIL — task_A12
14 actions traced
CAD inspected
deterministic tool
Connection candidates detected
deterministic tool
Surface inventory inspected
deterministic tool
Interface candidates evaluated
evidence-based decision
Engineer confirmation
source / target confirmed
Route generated
deterministic tool
Validation complete
PASS
Working Prototype
The current RouteForge prototype can execute an end-to-end workflow from CAD inspection through connection discovery, requirement resolution, route generation, engineering validation, and auditable output.
routeforge — task runner
Task complete
audit trail attached — every action, evidence record, and confirmation traced
GENERATED ROUTE — DN100
STATE VALIDATED
Prototype workflow visualization — a single illustrative run, not benchmark averages.
Honest Scope
RouteForge separates demonstrated capability from direction. The left column runs today in the prototype; the right column is where the platform is headed.
Direction, not commitments — no dates attached.
Use Cases
RouteForge is aimed at teams who turn equipment, requirements, and CAD into engineered designs — EPCs, skid and equipment manufacturers, process engineering companies, and plant engineering teams.
Automate connection interpretation and piping route development around packaged equipment.
Generate and validate routes between industrial equipment interfaces.
Reduce repetitive CAD-to-design workflow effort while preserving engineer authority.
Turn engineering requirements and geometry into repeatable, auditable workflows.
Assist engineers in evaluating routing and design changes in existing environments.
Why RouteForge
The system starts from physical CAD evidence — measured surfaces, dimensions, and topology, not assumptions.
Insufficient evidence causes investigation or escalation — not fabrication.
Engineering calculations and geometry operations are handled by deterministic capabilities.
Actions, evidence, decisions, confirmations, and validation results remain traceable.
Current Development
Built for the next generation of autonomous engineering.
RouteForge AI’s current prototype is being developed around the AI-BOOST Challenge 2 problem space: Agentic AI for Automated CAD Generation and Autonomous Simulation.
Our work explores how an engineering agent can autonomously inspect CAD, gather evidence, plan deterministic engineering actions, generate design geometry, validate the result, and involve an engineer only when necessary.
AI-BOOST Challenge 2 — Agentic AI for Automated CAD Generation & Autonomous Simulation
Challenge participation does not imply endorsement, funding, or selection.
The Bigger Vision
Engineering software has become excellent at helping humans create designs. The next step is software capable of participating in the engineering process itself — understanding geometry, gathering evidence, planning work, generating alternatives, validating results, and escalating judgment when necessary.
RouteForge AI is being built toward that future.
TODAY
Autonomous Piping
Interface discovery, routing, validation, and audit for industrial piping.
NEXT
Industrial Design Automation
More design domains, richer standards, requirements-to-design workflows.
VISION
Autonomous Engineering Platform
An engineering layer that understands geometry, plans work, generates and validates designs, and escalates judgment.
Building the Autonomous Engineering Layer
RouteForge AI is building infrastructure for autonomous industrial design — starting with piping.
Contact RouteForge AI
Direct contact details will be published here ahead of release. For engineering, partnership, or evaluation conversations, reach out through the channel where this site was shared with you.