Resume · Meshcapade

Turned graphics and machine-learning research into platform features

3D Graphics and Research Engineer · Tübingen, Germany · 2021 – 2023

A design proposal for defining surface features in UV coordinate space (July 2022). Meshcapade’s features (body segments, measurements) were defined as sets of vertex indices on its standardized body model, which made them fragile: lose the vertex order and a segment selection turns to geometric gibberish; change the model’s resolution and they break. A similar dependency on edge indices had already been removed after Blender versions reordered edges and broke the measurements code. The proposal: define surface features in UV space instead, so they survive retargeting to other topologies and resolutions.

A body-segment selection after vertex order is lost

Measurements: a Blender visualization of the platform’s body measurements, one render per measurement (height, waist, hip, chest, neck-base and thigh circumferences, arm lengths, shoulder breadth, foot length and more), each drawn as a line on the body mesh.

Waist circumference

Avatar profiles: aligning Meshcapade’s body template with Unreal Engine’s, comparing the two meshes and armatures (Meshcapade in T-pose, Unreal in A-pose), then aligning Meshcapade’s mesh and armature to Unreal’s.

Meshcapade's template (left) and Unreal's (right)

The two armatures overlaid after alignment

Asset Pipelines · Machine Learning · Graphics

DVC · Python · Blender · React