Project Overview
We are sourcing independent 3D Modeling and Python Specialists to provide their expertise for a specialized AI 3D asset collection project. As AI models increasingly generate and interact with 3D environments, their accuracy relies entirely on robust, expert-crafted training data. The objective of this project is to autonomously produce collision-accurate 3D meshes from 2D reference images, write procedural generation scripts, and deliver structured programmatic outputs (JSON/JS) to train AI systems on geometric construction and trajectory logic.
Project Deliverables & Scope
Operate autonomously to generate precise 3D assets and programmed outputs. Expected deliverables include:
Procedural Geometry Generation: Writing Blender Python (bpy) scripts or Three.js/JavaScript code to procedurally generate 3D geometry using geometric primitives and Boolean operations (union, subtraction, extrusion).
Mesh Alignment & Export: Producing collision-accurate meshes from 2D reference images (masked + RGB), capturing accurate screenshot alignments, and exporting assets to .obj, .glb, or similar formats.
Structured Output & Trajectory Logging: Producing programmatic JSON/JS outputs representing the mesh and maintaining meticulous trajectory logs that capture sequential build actions for reproducibility.
Quality Assurance & Workflow Adherence: Auditing all assets against strict quality checklists before submission, maintaining precise file naming conventions, and sustaining high-volume output within a structured annotation platform.
Required Expertise
To successfully fulfill the deliverables of this project, Contractors must possess a strong blend of 3D modeling fundamentals and programmatic scripting skills.
Core skillset includes:
Demonstrable proficiency in Blender Python scripting (bpy) and/or Three.js/JavaScript. Note: You must be able to write code to generate outputs; GUI-only modeling experience will not qualify for this project.
Deep understanding of geometric primitives (cubes, spheres, cylinders, cones) and Boolean operations, with a focus on low-poly construction rather than high-fidelity, subdivision-heavy sculpting (e.g., ZBrush).
Proven experience creating collision-accurate 3D meshes from 2D reference images.
Highly detail-oriented with the discipline to maintain trajectory logs, strictly follow file naming conventions, and self-audit work to maintain high reliability scores.
Preferred Experience: Background in game asset creation, familiarity with screen recording/keylogging tools, or prior experience within AI/ML data annotation workflows.
We offer a fixed pay rate of $22 per hour. As a contractor you’ll supply a secure computer and high‑speed internet; company‑sponsored benefits such as health insurance and PTO do not apply.
Engagement Type: Freelance / Independent Contractor
Workplace Type: Remote
Invisible Technologies is an AI training and data collection company that develops high-quality training data to enhance large-scale language models. The company specializes in recruiting domain experts—including general professionals, medical specialists, and language experts—to evaluate and improve AI system performance through structured feedback and testing. Their mission is to make tomorrow's AI smarter, more reliable, and more useful by leveraging human expertise to strengthen model reasoning across diverse fields including healthcare, linguistics, and general knowledge. They work with contractors to provide audio, video, image, and written inputs that support the development and refinement of cutting-edge AI systems.
