Parametric Workflow Automation
Custom Grasshopper and Python workflows that remove the repetitive work between teams and their next design decision.
What I do
Most simulation teams lose time to model preparation, data wrangling, and reporting, not the simulation itself. I build the pipelines that close that gap: parametric model setup, automated post-processing, ML surrogates where the physics engine is too slow, and reporting tools that hand back the right artefact to the right audience.
- Parametric simulation pipelines in Grasshopper, Ladybug Tools, Honeybee, TRNlizard, and TToolbox.
- Geometry preparation for CFD, radiation, and thermal models: including LiDAR2Building for instant LOD2 context geometry.
- ML surrogates to replace slow simulation engines for early-stage exploration (CNN daylight, GNN overheating, FFN peak solar).
- Cloud deployment on Azure ML and Rhino Compute with FastAPI, Docker, and GPU acceleration.
- Reporting tools: Plotly dashboards, Streamlit web apps, automated documents.
How I engage
- Workflow audit + targeted automation: find the 80/20 in a team’s existing process, automate the bottleneck.
- Tool build: bespoke Grasshopper component, Python library, or web app for a specific recurring need.
- Production ML pipeline: dataset generation, model training, deployment, and integration into the design workflow.
Methods
Python (Pandas, NumPy, FastAPI, PyTorch), Grasshopper with GH-Python and C#, Azure ML, Docker, Git. Tools are built to be handed over: documentation, version control, and team training included.
Deliverables
Grasshopper definitions, Python libraries, web apps, deployed ML endpoints, training documentation. Teams use the work, not just the report.
Example contexts
- LiDAR2Building: open tool, used on 20+ projects.
- Wind comfort automation for City of London (Grasshopper + Python + ANSYS).
- Outdoor comfort interactive reporting: Streamlit + Plotly on Heroku.