GIS Developer, Python, AI/ML, Mapbox, Routing, Logistics, ArcGIS

Job Description: Project Overview We are looking to build the Minimum Viable Product (MVP) for a predictive property intelligence platform for the UK market. The goal of the platform is to ingest fragmented public planning documents and property records, analyze them using NLP, and output predictive metrics (like a "Planning Upside Score") to an interactive map interface for investors. Scope of Work / Technical Requirements The MVP development is divided into three core pillars: 1. Data Ingestion & GIS Pipeline (The Core) o Set up automated scripts/scrapers to query and ingest data from UK sources (HM Land Registry, Open Digital Planning APIs, and local council planning portals). o Integrate a geospatial database (PostgreSQL / PostGIS) to handle land parcels, coordinates, and polygon boundaries. o Map geographic/policy relationships using a graph framework (e.g., Neo4j or relational equivalent) to link properties to local constraints. 2. NLP Engine (The Document Parser) o Build a pipeline using specialized NLP tools (e.g., Legal-BERT, Hugging Face Transformers, or custom LLM API pipelines via LangChain / LlamaIndex) paired with OCR (Tesseract / AWS Textract) to scan unstructured UK council planning PDFs. o The engine must automatically classify legal constraints (Permissive vs. Restrictive) and extract key entities. 3. Backend Logic & Frontend Dashboard o A Python-based backend (FastAPI preferred) implementing basic predictive math and machine learning libraries (XGBoost/LightGBM). o A clean, responsive frontend (React or Vue.js) featuring an interactive map using Mapbox GL JS or Leaflet, where users can click a land parcel and view its predictive scores. Required Skills & Tech Stack • Backend & GIS: Python, PostgreSQL/PostGIS, Neo4j, GDAL/Shapely • ML/NLP: Hugging Face, LangChain/LlamaIndex, Scikit-Learn, XGBoost • Frontend: JavaScript/TypeScript, Mapbox GL JS or Leaflet, React/Vue • Infrastructure: AWS or GCP, Docker Project Scope & Scope of MVP This is a phased project. For Phase 1 (this contract), we are limiting the scope to a single major UK geographic region (e.g., Leeds/West Yorkshire) to prove the pipeline works. The pipeline must successfully process data for 500+ sample properties and display them correctly on the map.