Unreal Engine 5 Developer

Wirestock is a leading data platform for ethically sourced multimodal data, providing licensed training datasets to AI labs. The Unreal Engine 5 Developer will independently build and optimize a technical rendering and data-export pipeline for producing richly annotated synthetic video datasets, including metadata extraction, derived annotations, and validation tooling for machine learning use.

Responsibilities:

• Implement and optimize render passes in UE5 (RGB, depth, optical flow, normals, amodal masks, and custom AOVs) against a locked metadata specification
• Build Python tooling to drive rendering, extract per-frame/per-object metadata, and export it into structured formats (parquet/JSONL), not just automate the editor
• Handle coordinate system and convention correctness (handedness, units, rotation representation, depth/flow semantics) so exported data is usable without buyer-side guesswork
• Implement derived/computed annotations (e.g., visibility episodes, occlusion attribution, articulation state) from raw render/transform data
• Build validation tooling (consistency checks, reprojection checks, overlay visualizations) to confirm exported data is correct
• Work independently against a detailed spec, flagging technical risk or scope tradeoffs as they come up

Requirements:

• Unreal Engine 5
• Render optimization (performance tuning for high-fidelity, multi-pass renders)
• Python scripting within UE5 (Editor scripting, automation, pipeline integration)
• Level Instances
• Chaos Physics
• Custom AOVs (Arbitrary Output Variables) / custom render pass authoring
• Python for data work. Comfortable with numpy/pandas-style structured data handling, not just script automation
• Experience working with EXR or similar multi-channel image formats as data (not just for compositing/viewing)
• Solid grasp of 3D coordinate transforms: quaternions, matrices, handedness/axis conventions, and converting between them
• Comfort designing a data schema and export pipeline at scale (structured tabular formats, consistent file/frame organization)
• Git
• Prior experience producing data for ML/computer vision training (synthetic data, sim-to-real, robotics, photogrammetry, or similar)
• Experience with data export pipelines that were later consumed/analyzed outside the engine (e.g., in Python/Jupyter, a CV pipeline, or a data science workflow)
• Movie Render Queue
• PCG (Procedural Content Generation) framework