ParaView · VTK · Catalyst · CFD & FEM post · in-situ HPC
A decade across ParaView, VTK, Catalyst, CFD and FEM post-processing. Most of my pipelines are scripts because a state file you can diff is a visualization you can reproduce.
render_figures.pyFive projects that together cover the scientific-visualization stack — from raw simulation data through interactive analysis, custom filters, and in-situ pipelines that run inside the solver.
CFD post · vortex identification · VTK
The reference workload for any CFD post-processing pipeline. A 64³ analytic vortex field, a Q-criterion derived from velocity gradients, an isosurface for vortex cores, and RK4-integrated streamlines tubed and colored by speed. Every step is scripted — .vti output drops straight into ParaView.
FEM post · unstructured mesh · warp-by-vector
End-to-end FEM-to-ParaView workflow: build a tetrahedralised notched beam (~78k cells), attach a mechanically consistent stress tensor with a notch concentration, derive the von Mises invariant per point, and render both the static field and the warped deflected shape. Includes the percentile clip every junior engineer learns the hard way.
Read the case study →
Volume rendering · transfer functions
96³ procedural plume, exported to .vti, rendered with a hand-tuned color + opacity transfer function. Companion isocontour view at three nested levels for quantitative readability.
Plugin · vtkPythonProgrammableFilter
Server-manager XML proxy plus Python module that adds a single new filter: compute vorticity on any vector array on any topology, and emit the top-N highest-vorticity points as a vtkPolyData ready for Stream Tracer With Custom Source.
Catalyst v2 · in-situ · HPC
A toy time-stepping driver and a Catalyst v2 adaptor that contours, color-maps, and screenshots in-process — the strategy whenever the solver produces more data per step than disk can hold.
Case study →Every figure on this page is regenerated by GitHub Actions on every push — never pre-baked. No proprietary data, no commercial solver licenses required. Clone the repo, install VTK, get the same picture.
python scripts/python/taylor_green_vortex.pyBuilds the 64³ velocity field, derives Q-criterion, writes .vti, renders the isosurface and streamlines.
python scripts/python/fem_stress_field.pyTetrahedralises the beam, attaches σ + displacement, writes .vtu, renders the von Mises view and the warp-by-vector view.
python scripts/python/volumetric_scalar.py96³ procedural plume, GPU volume render with a custom opacity ramp, plus an isocontour view at three levels.
python scripts/python/render_figures.pyRebuilds every PNG that ships in /docs/assets — what GitHub Actions runs on every push.
Daily driver since 4.x. State files, pvbatch, pvpython, multi-view layouts, custom transfer functions, programmable filters and sources.
C++ and Python. Image data, structured/unstructured grids, polydata, vtkAlgorithm subclassing for custom filters. Headless OSMesa rendering for CI.
Server-manager XML proxies, Python plugins, C++ plugins via vtkPVPluginLoader. Shipped plugins to Slurm clusters and Windows workstations.
Catalyst v2 adaptors with Conduit, Catalyst Live for steering. ParaView client–server over MPI. GPU offload via OSMesa / EGL.
OpenFOAM, SU2, CGNS, EnSight Gold, XDMF, Exodus II, NetCDF, HDF5. Ensemble post-processing and time-series analysis.
Python (numpy, scipy, vtk, paraview), C++17. Git, GitHub Actions, Docker. Every analysis is a script — diffable, reviewable, reproducible.
I'm a senior scientific-visualization specialist with ten years across ParaView, VTK, and Catalyst, doing CFD and FEM post-processing and in-situ visualization for HPC simulations. I care about reproducible pipelines, transfer functions that don't lie, plugins that ship cleanly, and visualizations that still render the same way a year from now.
Open to remote and contract engagements in scientific visualization, CFD/FEM post-processing, or HPC in-situ workflows. The repository linked below is the living portfolio companion to my CV.