Computer Vision AI Engineer for Non-Verbal Reasoning Figure Reconstruction & Generation

I am looking for an experienced Computer Vision / AI Engineer to build a reusable pipeline for reconstructing and generating Non-Verbal Reasoning question images used in competitive examinations. This is not a simple image generation project. We are looking for someone who can build a scalable solution that accurately understands, reconstructs, and eventually generates examination-quality figures. Objective 1 – Exact Reconstruction of Existing Figures (PoC) Develop an end-to-end pipeline that reconstructs existing/reference non-verbal reasoning question images with near-identical accuracy. The reconstructed images should preserve: Geometry Relative positioning Spacing Alignment Rotation and orientation Proportions Line styles Symbols Patterns Overall visual appearance The output should be clean, production-ready SVG (preferred) and PNG images. The PoC will be evaluated using a representative dataset of sample questions. Objective 2 – AI-Assisted Generation of New Non-Verbal Reasoning Figures The long-term goal is to use the same architecture to generate new non-verbal reasoning questions. Given a reference question, the solution should be extensible enough to eventually generate: A new question image Corresponding option images Explanation image while maintaining: The same question pattern The same reasoning concept The same visual style The same difficulty level expected in competitive examinations The objective is to help us build a large database of new non-verbal reasoning questions that currently require significant manual effort using tools like CorelDRAW. Note: The initial engagement (PoC) focuses on Objective 1. However, the proposed architecture should be reusable and extensible so that Objective 2 can be implemented in a future phase without redesigning the entire system. Preferred Architecture We prefer a reusable architecture rather than a one-off solution. Ideally, the pipeline should: Convert reference images into a structured intermediate representation (JSON or equivalent). Use the structured representation to generate editable SVG and PNG outputs. Include a reusable rendering engine that can support future AI-assisted figure generation. Be modular and suitable for API-based integration into our existing platform. Deliverables Working end-to-end reconstruction pipeline. Reconstruction of the provided sample dataset. Structured intermediate representation (JSON or equivalent). Reusable rendering engine. SVG and PNG outputs. Complete source code. Documentation explaining the architecture and setup. API or callable interface for future integration. Required Skills Computer Vision Image Processing OpenCV (or equivalent) Python SVG / Vector Graphics OCR Geometry Processing Vision LLMs / Generative AI (preferred) Nice to Have Experience with: Diagram understanding Image-to-vector reconstruction Educational or assessment content AI-based document/image processing Rendering engines Multi-agent AI workflows