iOS Multi-Camera Photogrammetry Engineer - Sub-mm 3D Reconstruction

Looking for a senior engineer with real hands-on experience in multi-view geometry, camera calibration, and photogrammetric 3D reconstruction to build out an iOS capture app that produces sub-millimeter accurate 3D models of small objects. This is NOT a wrapper around Polycam, Object Capture, or Kiri SDK. If sub-mm accuracy isn't achieved on our target object, the work has no value to us. What's already built We have an existing iOS wireframe app that captures from the front (TrueDepth) and rear (LiDAR) cameras and exports OBJ / PLY. That's the starting point. We're not looking for another depth-sensor build - that direction has been ruled out on accuracy grounds. What we need Rebuild the capture pipeline around synchronized multi-camera stereo capture using the iPhone Pro rear array (Ultra-Wide, Wide, Telephoto) via AVCaptureMultiCamSession. Use the known rigid baseline between the three lenses as a calibrated stereo rig. Send synchronized frame clusters plus intrinsics and extrinsics to a cloud reconstruction pipeline (FastAPI backend already in place). On the server: SfM with rigid-rig bundle adjustment (fixed relative extrinsics), scale fixed from the physical camera baseline. Output: watertight mesh in OBJ / PLY. Hard requirement: the pipeline must demonstrably achieve better than 1mm accuracy on a physical test object we'll provide. If it doesn't hit that, the deliverable isn't complete. Scope Extraoral only. Bench-scale physical objects, not intraoral. Sub-mm on the target object is the whole point. How to apply Hourly, submit your rate and an honest estimate of hours to complete. In your proposal, please answer: 1. Link to prior work involving multi-view stereo, SfM, or photogrammetric reconstruction. Papers, GitHub, shipped products. 2. What's your approach to calibrating T_rig between iPhone Pro rear cameras? Factory intrinsics only, or offline calibration target? 3. What's the fundamental limit on depth accuracy given the ~15mm baseline between iPhone Pro Wide and Telephoto at 20cm working distance? 4. Have you worked with COLMAP's rig extension, or OpenMVG / OpenMVS? Which and in what context? Interview Shortlisted candidates get a live video technical interview. LLM-generated proposals are easy to spot; a real conversation isn't. If you can't discuss epipolar geometry, feature matching, bundle adjustment, and scale ambiguity live without slides or notes, don't apply. No agencies. Direct communication with the developer only.