Computer Vision Engineer, VIO / SLAM

About the company

Efference provides high-performance perception and compute infrastructure for training and deploying robot models in the cloud. Our vision is to make robotic deployments scalable with shared compute.

We approach the problem by owning the entire path from the image sensor to the datacenter GPU. We shrink the load at the sensor, fix the uplink so robot traffic is fast and reliable, and clock synchronize the system so observations arrive time-aligned, ready for inference.

About the role

You'll develop Efference's visual-inertial odometry and mapping stack, including the calibration that makes its outputs trustworthy. You'll combine geometric estimation with learned components where they help, deploying the system on our embedded compute and measuring how imaging and compression choices affect localization.

What you’ll do

• Build visual-inertial estimation covering feature tracking, initialization, pose estimation, sensor fusion, relocalization, and mapping for real robot motion.
• Develop camera and IMU intrinsic, extrinsic, and temporal calibration, with procedures that can be used during development and manufacturing.
• Investigate classical, learned, and hybrid front ends; optimize suitable learned components for our NPU while maintaining the estimation pipeline's accuracy and timing.
• Quantify sensitivity to exposure, blur, rolling shutter, ISP processing, frame loss, compression, and synchronization error with the camera and wireless teams.
• Build benchmarks for trajectory error, drift, tracking loss, recovery, and runtime using recorded sequences and physical ground truth.
• Ship maintainable C++ implementations with calibrated coordinate frames, timestamps, uncertainty estimates, and diagnostics for downstream robotics software.

What we are looking for

• Substantial experience developing VIO, SLAM, visual odometry, or multi-sensor state estimation.
• Strong understanding of multi-view geometry, camera models, nonlinear optimization, and probabilistic estimation.
• Experience with IMU measurements and camera/IMU calibration, including the effect of timing and noise models on observability.
• Strong C++ and Python skills and a record of evaluating algorithms on real sensor data.
• Ability to diagnose tracking failures and improve both the algorithm and the surrounding sensor pipeline.

Nice to Have

• Experience with learned feature extraction, optical flow, depth, neural odometry, or embedded NPU deployment.
• Work with fisheye or global-shutter cameras and libraries such as Ceres, GTSAM, Eigen, or OpenCV.
• Experience with factory calibration, motion-capture ground truth, or VIO on legged and mobile robots.