AI for Engineering
PRACTICE / 01Forward Deployed AI Engineer · Engineering Systems Architect
I connect physical engineering, software architecture and applied AI to turn fragmented information into structured, traceable and reviewable systems. My experience spans AEC, BIM/VDC, electrical systems and critical infrastructure.
physical systems computational models engineering intelligence reviewable results
The handoff
Good engineering software does not replace context with a confident answer. It keeps the object, its references and the decision connected long enough to inspect what changed.
Enter the signature system →The signature system
AI for Engineering is not about adding AI to a workflow. It is about building software that understands the work, acts inside it and improves with evidence.
Reality
The work starts with equipment, spaces and constraints. I build software for engineering work that has to hold up in the real world.
Every representation that follows still has to describe this same physical system.
Forward Deployed AI Engineer · Engineering Systems Architect
Model
The project appears through models, drawings and schedules.
Each representation reveals something useful and hides something else.
Context
The same object needs the same identity across the model, drawing, schedule and requirement.
Once identity is stable, a question about an engineering entity can reach its model representation, source records and requirement without guessing which object each source means.
AI-assisted analysis
With the references aligned, the agent can trace one question across the model, drawing and schedule.
Once the references line up, the agent can follow a question from requirement to model, drawing and schedule without losing where each observation came from.
Agentic workflow
A bounded agentic workflow turns one question into a sequence: inspect structured data, follow source links, then look for contrary evidence.
The workflow searches structured data, follows sources and checks contrary evidence — inside explicit authority limits.
Illustrative verification example; these values are not measurements extracted from the displayed model.
Illustrative action trace: inspect model, query structured data, open a source record or search contrary evidence. A source record supplies the next observation.
Explore the work ↗Verification
The source record carries the decisive observation: in this illustrative check, one representative element falls below a stated condition.
A single counterexample is enough to stop the claim. The sheet, element and rule remain visible together.
Illustrative verification example; these values are not measurements extracted from the displayed model.
Illustrative requirement: an engineering attribute must meet a stated threshold. One representative element falls below it in this synthetic example. These values are not extracted from the displayed BIM.
Explore the work ↗Feedback
Change the workflow, observe what happened, carry that signal into the next run.
A result becomes useful when it changes an action, creates a record and can be checked again on the next run.
Illustrative verification example; these values are not measurements extracted from the displayed model.
Illustrative output: a requirement is blocked, with one representative counterexample requiring attention. Create an issue, assign a reviewer, open the element, re-run and track resolution. The report preserves the requirement, evidence, rule, review status and run state.
Explore the work ↗Drag to rotate · pinch to zoom · Esc to continue
Selected work
Connecting requirements, models, schedules and drawings into a traceable workflow for analysis, validation and human review.
Read the approach →
Production add-ins designed around a stateful host, multi-version compatibility, transactions, model change and release reality.
Explore the integration work →
Professional scope
Solar taught me to discover, engineer, sell and deploy in the physical world. BIM made that world computational. Mission-critical work raised the consequence. Software increased the leverage. Trace the full path →
Independent computational research
Independent research explores spectral operators, structure-preserving algebra and computational models with explicit claim status: definition, proof under assumptions, numerical verification, conjecture or open.