About Mecka AI

Mecka AI is building the data infrastructure layer for robotics and embodied AI.

We design and operate global systems for data capture, data labeling, and hardware-enabled workflows used by leading AI labs and robotics companies to train and validate humanoid and embodied AI systems.

We work closely with frontier robotics teams to bridge real-world data, simulation, learning-based systems, and deployed hardware.

The Role

We're looking for a hands-on Senior SLAM & Calibration Engineer to own how Mecka's devices stay calibrated after they leave the factory. Cameras and IMUs drift over months and mounts warp, so a device that shipped in spec can fall out of it. You'll build the systems that catch and correct that drift: online (in-pipeline) camera calibration, video-based calibration refinement, fleet-wide drift monitoring, and the tools operations and researchers use to track it, all built on solid SLAM/VIO foundations.

This is a production role, not pure research: you'll own systems that run across the real fleet. You'll work directly with the CVML team and coordinate with our factory-calibration engineer and hardware team. Because the work runs on real, hardware-synced recordings and is validated on physical devices, this is an on-site role with periodic travel to our hardware site.

What you'll do

Own online calibration: Develop and maintain continuous, in-pipeline camera calibration that refines extrinsics from recorded, hardware-synced, uncompressed video and IMU — the pipeline equivalent of on-device calibration, so fixes ship without a device software update.

Rebuild video-based calibration: Fix and harden the video-based calibration refinement system across the fleet, including the monocular wrist-cam path, and drive it against real benchmarks.

Monitor the fleet: Build the monitoring and metrics tracking that detects calibration drift across every device, flags devices for recall, and owns the calibration database and version history.

Make SLAM carry calibration: Maintain and improve the SLAM/VIO that online calibration rides on, and have it emit calibration-related error as a first-class output.

Bridge to hardware: Coordinate with the factory-calibration engineer and the China hardware team on the intrinsics limit (online refinement fixes extrinsics; intrinsics need a marker at the factory), IMU noise and bias parameters, temporal synchronization, and validation runs that avoid local minima.

Build the tooling: Ship interactive tools (Rerun / Gradio) that visualize trajectories, drift over time, reprojection error, and per-device calibration metrics for operations and researchers.

What You Bring

Must-have:

SLAM / VIO / SfM depth: 5+ years of hands-on experience building, maintaining, or improving SLAM, visual-inertial odometry, and Structure-from-Motion pipelines, including systems running in production.

Continuous and multi-sensor calibration: Hands-on experience with online or continuous calibration and rigorous camera + IMU calibration: extrinsics, intrinsics and their limits, IMU noise density and random-walk bias, lens-distortion models, and temporal synchronization.

Fleet and data-quality mindset: You're comfortable owning calibration across many noisy, real-world devices, and you build the monitoring that catches drift before customers do.

Multi-view geometry and optimization: An intuitive grasp of the linear algebra, optimization, and first principles behind spatial tracking.

Engineering rigor: Clean, efficient, scalable C++ and Python.

Cross-border collaboration: You can specify calibration requirements clearly to a hardware team across time zones.

Nice-to-have:

Markerless calibration: Experience with video-based or markerless calibration and refinement systems.

Calibration frameworks: Sensor-fusion and calibration tools such as Kalibr, GTSAM, or Ceres Solver.

3D vision and ML libraries: OpenCV, COLMAP, PyTorch, and FFmpeg.

Fleet observability: Monitoring at scale, including calibration databases and versioning.

Spatial tooling: Rerun, Gradio dashboards, or trajectory and dataset browsers.