At Luxolis, we don't just build cameras; we build Spatial Intelligence. We are moving beyond simple 2D imaging to create a world where machines have a human-like understanding of 3D space.
We are seeking a 3D Spatial Vision Intelligence Engineer to lead the development of our proprietary Spatial Vision Chip. This is a rare role for an engineer who thrives at the intersection of SLAM (Simultaneous Localization and Mapping) and High-Performance Hardware Acceleration. You will be responsible for hardcoding spatial awareness into silicon, enabling real-time, low-latency 3D reconstruction for industrial robots and digital twins.
What You Will Do- Architect the Vision Chip: Design and implement FPGA-based architectures to accelerate core spatial tasks: depth estimation, feature tracking, and point-cloud processing.
- Hardware-Accelerated SLAM: Port, optimize, and harden state-of-the-art SLAM algorithms (e.g., ORB-SLAM, VIO, LOAM) from C++ into RTL (Verilog/VHDL).
- Sensor Fusion at the Edge: Develop high-speed data pipelines to fuse multi-modal inputs (LiDAR, Stereo, IMU) directly on the FPGA to achieve sub-millisecond latency.
- AI-Hardware Co-Design: Collaborate with our AI team to optimize 3D neural networks (like PointNet++ or 3D Segmentation) for deployment on custom hardware.
- RTL to Reality: Manage the full FPGA lifecycle—from HLS prototyping and RTL coding to timing closure, verification, and on-device deployment.
Who You Are- The Bridge: You can read a SLAM research paper and visualize exactly how those matrix operations should be pipelined in hardware.
- FPGA Power User: Mastery of Verilog/VHDL and toolchains like Xilinx Vivado or Intel Quartus. Experience with HLS (High-Level Synthesis) is a major plus.
- 3D Geometry Expert: Deep understanding of the math that powers vision—quaternions, transformation matrices, epipolar geometry, and Kalman filters.
- C++ Expert: Strong proficiency in modern C++ and experience with libraries like OpenCV, PCL, or Eigen.
- Experience: 3+ years in computer vision or robotics, with a proven track record of deploying algorithms on embedded hardware or FPGAs.
To stand out for the 3D Spatial Vision Intelligence Engineer role at Luxolis, candidates must possess a Full-Stack Hardware mentality. This means combining the high-level mathematical theory of 3D geometry with the low-level constraints of RTL and digital logic.
Below are the detailed qualifications categorized by priority.
Educational Background
- Minimum: Bachelor's or Master's degree in Electrical Engineering (EE), Computer Engineering (CE), Robotics, or a related field.
- Preferred: A PhD with a research focus on Computational Imaging, SLAM, or Hardware Acceleration for Computer Vision.
- Key Coursework: Digital Logic Design, Computer Architecture, Linear Algebra, Probability & Stochastic Processes, and Computer Vision.
Technical Skills: The Spatial Intelligence Stack
These are the core competencies required to design a chip that sees and understands 3D space.
1. SLAM & 3D Perception- Algorithmic Expertise: Deep understanding of Visual-Inertial Odometry (VIO), LiDAR SLAM (e.g., LOAM, LeGO-LOAM), and Visual SLAM (e.g., ORB-SLAM3).
- Geometric Math: Mastery of SO(3) and SE(3) Lie Groups, quaternions, epipolar geometry, and bundle adjustment.
- State Estimation: Hands-on experience with Extended Kalman Filters (EKF), Unscented Kalman Filters (UKF), and Factor Graph Optimization (GTSAM/Ceres).
2. FPGA & RTL Engineering- Hardware Description Languages: Expert-level Verilog or SystemVerilog (preferred) or VHDL.
- Design Tools: Proficiency in Xilinx Vivado, Intel Quartus, and High-Level Synthesis (HLS) for rapid algorithm prototyping.
- Architectural Knowledge: Experience with AXI protocols, DMA, and memory controllers (DDR4/HBM) to handle high-bandwidth 3D point cloud data.
- Verification: Experience with UVM/OVM or hardware-in-the-loop (HIL) testing to ensure first-time-right silicon logic.
3. Software & Optimization- Primary Languages: High proficiency in Modern C++ (14/17/20) for performance-critical systems and Python for algorithm modeling.
- Vision Libraries: Strong experience with OpenCV, PCL (Point Cloud Library), and Eigen.
- Acceleration: Familiarity with SIMD instructions or CUDA (for benchmarking FPGA performance against GPU equivalents).
