Mid-Level Computer Vision & 3D Deep Learning Engineer

Job description

About the role: We are looking for a Computer Vision Engineer with a solid background in deep learning and 3D data processing to join our team. You will work on developing and deploying models that understand and reconstruct the visual world, contributing to production-grade pipelines that take multi-view 2D images and produce high-quality 3D reconstructions (from statistical shape models to implicit neural representations and texture synthesis). at the intersection of classical 3D geometry and modern neural approaches. This role is ideal for someone with 2–3 years of hands-on experience who enjoys bridging research and production, and is comfortable designing and training pipelines, evaluating reconstruction quality, and integrating your work into a complex multi-stage system. Responsibilities Research, prototype, and integrate new deep learning algorithms from recent literature (NeurIPS, CVPR, ICCV, ECCV) to improve 3D reconstruction quality. Develop and maintain deep learning components for multi-view reconstruction, landmark detection, segmentation, inpainting, and view-consistent shape fitting. Implement and tune custom training pipelines and loss functions, and evaluate their impact on mesh and texture quality. Design and run quantitative evaluation experiments using metrics such as reprojection error, surface-to-surface distance, and perceptual quality scores Export and deploy trained models for inference (TorchScript/JIT, Triton Inference Server,..) Our ideal candidate would have: 2–3 years of hands-on experience in computer vision and deep learning research or applied engineering Solid understanding of camera models, projective geometry, and multi-view geometry (epipolar geometry, camera calibration, reprojection) Experience training and debugging neural networks end-to-end, including custom loss functions, learning rate scheduling, and training stability Comfortable reading and implementing methods from academic papers Strong Python skills; proficiency with PyTorch (primary) and/or TensorFlow Comfortable working in a research codebase with complex multi-stage pipelines Fluent or proficient in English (Spanish is a plus). We also value very positively: Experience with 3D vision techniques (e.g. NeRFs, differentiable rendering, SLAM). Understanding of implicit surface representations: Signed Distance Functions (SDFs), occupancy networks, NeRF/neural radiance fields Familiarity with classical 3D fitting approaches: statistical shape models (PCA-based), iterative closest point (ICP), mesh deformation Knowledge of differentiable rendering concepts: ray marching, sphere tracing, volume rendering Familiarity with libraries such as Open3D, PyTorch3D, or OpenCV. Experience with experiment tracking tools (MLflow, W&B) and reproducible training pipelines Experience deploying models to production environments, using Docker to ensure reproducibility and scalability. Understanding of GPU optimization and performance tuning. Background in geometry, linear algebra, or graphics.