Jessy Leite

• Refactored the collision detection system for optimizing the positioning of frames on 3D facial scans, improving the glasses matrix and optical measurements (Three.js, TypeScript). Precision 2 → 0.5 mm (−75%), extended case coverage (complex geometries, extreme fits), performance maintained under 300 ms @ M1.

• Integrated ML-based landmark detection models to automate the 3D scan acquisition process (Three.js, TypeScript, MediaPipe). Support > 99.9% landmarks, robust 3D orientation on various scans, latency < 1 s @ M1.

• Developed advanced 3D features on assignment for Adobe on creative R&D projects (Python, Blender, Three.js).

• Initiated and developed a procedural auto-rigging methodology for humanoid characters (T-pose/A-pose), enabling skeleton and skinning generation without manual steps (Python, Blender). 95% coverage of the sample of 60 models in regression tests, average duration of 10s per model on M1), similar coverage to Mixamo (internal tests).

• Refactored Automap: replaced the naming convention approach with a combined topological (graphs) and geometric analysis, extending coverage of compatible rigs and eliminating naming dependencies (Three.js, TypeScript). +50–70% of estimated supported rigs, removal of naming constraints, increased topological robustness.

• Debugged and optimized animation retargeting systems for key accounts, resolving critical issues with compatibility between rig structures and 3D model formats (Python, Blender).

• Developed a SaaS platform automating the conversion of 2D vector drawings (SVG) into manufacturable 3D models (STEP/STL) for CNC machining and 3D printing, eliminating manual modeling for eyewear manufacturers (Three.js, React, TypeScript).