Build an on-prem Python platform for 3D property evidence tours + measurement + UK business rates Check & Challenge packs (PDF + portal)

We need a developer (or small team) to build an on-prem hosted web platform that turns survey capture files (including 3D Gaussian splats / scan outputs) into:
1. a customer/VOA-facing evidence portal (read-only URL) with an embedded interactive walkthrough (krpano), pins/notes, guided tour playback, and labelled evidence images; and
2. professional Check and Challenge PDF packs for the UK business rates “Check, Challenge, Appeal (CCA)” process, with measured floor plans, area schedules, zoning, exclusions, evidence appendix, and deep links back to the exact viewpoints inside the tour.

This must be built for England & Wales workflows, but the developer does not need business rates expertise: we provide a detailed brief, examples, and test cases. What we need is strong engineering, correctness, and auditability.

High-level user flow
• Surveyor uploads a large survey package remotely (resumable upload) to the on-prem server.
• System runs a processing pipeline to create derived assets (tour build, thumbnails, draft plan outputs).
• Assessor uses a Measurement Studio to draw/edit room polygons, zones, and exclusions (pillars/WC/plant/etc.) with live area totals and variance vs VOA baseline.
• Assessor creates pins and screenshots in the tour, labels them, and chooses whether they appear in Check, Challenge, or both.
• System evaluates the “scheme” (valuation model) and can simulate scenarios (deterministic calculators + trace).
• System generates:
• Check Pack PDF + Check URL
• Challenge Pack PDF + Challenge URL
• Published packs are immutable (versioned manifest): later edits create new versions/packs, but published packs never change.

Core requirements
• On-prem deployment (Docker Compose at minimum).
• Python backend: FastAPI preferred.
• Worker pipeline: Celery (or similar) for long processing tasks.
• Storage: S3-compatible object storage (MinIO) for uploads and generated assets.
• Database: Postgres + PostGIS for geometry (polygons).
• Tour viewer: krpano embedded in the app, with deep links to exact viewpoints.
• Evidence system: pins/hotspots, screenshots/figures with labels and inclusion flags.
• Plan editor: polygon draw/edit, split/merge, zoning tools, exclusions, live totals.
• Scheme engine: deterministic eligibility + RV simulations (we provide scheme database; you build the engine interface and calculators framework). AI can assist with drafting but must not invent facts; deterministic results must be traceable.
• PDF generation: Check and Challenge templates with plan sheets, schedules, variance tables, evidence appendix with QR/deep links.
• Portal sharing: scoped share tokens (read-only), expiry, access logs.

Deliverables
• A working codebase (repo + Docker deployment) that a developer can set up and run on-prem.
• Documentation:
• setup/runbook
• API docs
• developer notes
• QA “golden path” acceptance test steps
• Demo dataset and end-to-end demo case (“DEMO-SHOP-001”) that can be used for testing.

Milestones
1. MVP Golden Path: project → VOA snapshot → upload → pins/figures → generate Check PDF + Check URL portal
2. Measurement Studio v1: rooms/zones/exclusions + variance vs VOA + plan sheets export
3. Scheme Engine v1: eligibility + deterministic simulations + trace + scheme-change workflow (where eligible)
4. Challenge Pack: narrative + simulations + evidence appendix + Challenge URL portal
5. Hardening: RBAC, audit trail UI, upload reliability, backups/logging, performance

Acceptance criteria
• Golden path passes end-to-end on a clean deployment (fresh DB + storage).
• Published pack immutability: edits after publish do not alter existing pack.
• Deep links from PDF/portal open the exact viewpoint in the tour.
• Deterministic simulation outputs have a computation trace stored and viewable.

Engagement
• Prefer fixed-price per milestone with clear deliverables.
• Must commit code regularly, write tests where practical, and document deployment.

Please refer to the attached proposed user interface screens.

Please also refer to the PDF project brief, zip file of example code (created using ai) as a starting point. I have also included a backlog PDF to show what needs to be completed from there.