3D Computer Vision Engineer (3D Reconstruction)

DeepX is on a mission to push the boundaries of 3D perception. We’re looking for an experienced Computer Vision Engineer  (3D Reconstruction) who lives and breathes reconstruction pipelines, thrives on geometry challenges, and wants to architect systems that redefine how machines see the world.

This is not a plug-and-play role. You’ll be building the core reconstruction engine that powers our vision products - designing a modular, blazing-fast pipeline where algorithms can be swapped in and out like precision-tuned gears. Think COLMAP on steroids, fused with neural rendering, and optimized for scale.

    Core Tech Stack:

• Languages: C++, Python
• CV/3D Libraries: OpenCV, Open3D, PCL, COLMAP
• Math/Utils: NumPy, Eigen
• Visualization: Plotly, Matplotlib
• Deep Learning: PyTorch, TensorFlow

• Data: Point clouds, meshes, multi-view image sets.
   

  Desired Expertise: 

• Core Expertise: Deep, hands-on knowledge of 3D computer vision fundamentals, including projective geometry, triangulation, transformations, and camera models.
• Algorithm Mastery: Proven experience with point cloud and mesh processing algorithms, such as ICP for registration and refinement.
• Development Experience: Strong software engineering skills, primarily in a Linux environment. Experience deploying applications on Windows (or via WSL) is a major plus.
• Data Handling: Experience managing and analyzing the large datasets typical in 3D reconstruction.
  Projective Geometry Mastery: Camera models, projections, triangulation, multi-sensor fusion.
• Transformations: Rotations, quaternions, coordinate system conversions, 3D frame manipulations.
• SfM & MVS: Proven hands-on with pipelines and dense reconstructions.
• SLAM: Bundle adjustment, pose graph optimization, loop closure.
• Code Craft: Strong software engineering chops - designing modular, performant, production-grade systems.
• Visualization: Proficiency with 3D visualization tools and libraries (e.g., OpenGL, Blender scripting) for rendering and debugging point clouds and meshes.

• Bonus Points:
  - You’ve built a full 3D reconstruction pipeline from scratch.

  - Hands-on with Gaussian Splatting or NeRFs.

  - Experience with SuperGlue or other state-of-the-art feature matching models.

    - Hybrid reconstruction experience: fusing classical geometry      with neural methods.

   - Experience with real-time or streaming reconstruction systems.

   - Familiarity with emerging topics like 3D scene segmentation and the application of LLMs to geometric data.

What You’ll Do:

• End-to-End Pipeline Development: You will architect, build, and deploy a robust, high-performance 3D reconstruction pipeline from multi-view imagery. This includes owning and optimizing all core modules: feature detection, matching, camera pose estimation, SfM, dense stereo (MVS), and mesh/surface generation.
• System Architecture: Design a highly modular and scalable system that allows for interchangeable components, facilitating rapid A/B testing between classical geometric algorithms and modern neural approaches.
• Performance Optimization: Profile and optimize the entire pipeline for low-latency, real-time performance. This involves advanced GPU programming (CUDA/OpenCL), efficient memory management to handle large models, and leveraging modern compute frameworks.
• Research & Integration: Stay at the forefront of academic and industry research. You will be responsible for identifying, implementing, and integrating state-of-the-art methods in SLAM, neural rendering (NeRFs, 3DGS), and hybrid geometry-neural network models.

• Data Management: Develop solutions for handling, processing, and distributing large-scale image and 3D datasets (e.g., using tools like Rclone).
  
  Why Join DeepX?

  This is your chance to own a core engine at the frontier of 3D vision. You’ll be surrounded by a small but elite team, working on real-world deployments where your algorithms won’t just run in benchmarks - they’ll run in airports, mines, logistics hubs, and beyond. If you want your code to shape how machines perceive the world at scale, this is the place.

   

  Sounds like you? -> Let’s talk
  Send us your portfolio, GitHub, or projects - we love seeing real reconstructions more than polished CVs.