iOS Prototype — LiDAR or Meta Glasses Room Scanning + Label Recognition (paid trial first)

We are a European wine custody and advisory firm building a private inventory app for a small number of wine collectors. We are looking for one experienced developer to build a Phase 1 iPhone prototype, and we begin with a short paid trial before committing to the full build. The prototype has two halves that work as one flow. First, spatial capture: the user walks a wine cellar with an iPhone and captures its geometry using ARKit / RoomPlan or similar, producing a simplified, manually correctable 3D schematic of racks, shelves and fridges, with each rack subdivided into addressable slots. Second, bottle recognition: pointing the camera at bottles, the app reads labels with OCR and image recognition (on-device Vision plus a hosted vision or LLM service - no custom model training), matches each against a reference wine database we supply, returns a confidence score, and the user confirms hands-free by voice with touch fallback. Confirmed bottles are written to inventory with vintage, quantity and slot. The whole loop must work offline underground and sync later. It should be as fast and frictionless as possible. Ideally, it could also work on video mode or with Meta Glasses. Acceptance target for the full build: log a fifty-bottle rack in one continuous pass, correct identification on 85%+ of clean front-facing labels (minimum acceptable 80%), at most one touch per bottle in the failure case. Architecture requirement: the capture layer must be a swappable module behind a documented interface, with recognition, matching and inventory logic in a device-agnostic backend, so that a smart-glasses capture client can be added in a later phase without a rewrite. Please tell us what you know about current third-party access to Meta's display glasses. Paid trial task (6-8 hours, first engagement): we provide roughly twenty photographs of real bottles shot in actual cellar lighting. You return a working script or small app that OCRs them, matches them against a trimmed reference list we supply, and outputs a confidence score per bottle. We are looking at how you handle ornate, reflective labels in poor light. Whoever performs best is offered the full prototype on a milestone basis. Out of current scope: pricing and valuation of the wine bottles, wine recommendations, Android, web, account management. In your proposal, please answer: which scanning framework you would use and why; your approach to difficult label conditions; how you would run voice confirmation offline; estimated hours and weeks per milestone for the full build; whether you work solo or with an agency team; and one shipped app you have built involving camera-based recognition or spatial capture in a real environment. Budget: $18-40/hr depending on experience. NDA required before any data or imagery is shared. Please skip generic proposals - we read the answers above first.