Geospatial Engineer (Python / PostGIS) - Property Assessment Platform

We're building a platform that produces development-feasibility assessments for Australian residential properties - address in, planning constraints computed, assessment report out. We're scoping the engagement now. The core of it, and the part where we most need real expertise, is computing development constraints from parcel geometry and rendering them onto maps clear enough for someone to make a decision from. How far beyond that core the role extends is something we'd rather work out with the right person than fix in advance. The inputs are a cadastral parcel polygon, its attributes, and aerial imagery. From those, compute the buildable envelope after applying setbacks that vary by boundary type, then render it as a correctly georeferenced map suitable for a report. The challenge is variance rather than volume. Every parcel shape in the jurisdiction runs through the same code path unattended: corner blocks with two frontages, battle-axe blocks whose access handle isn't a frontage at all, irregular non-convex subdivisions, parcels with existing structures in awkward positions, and source geometry carrying the usual real-world slivers and invalid rings. A pipeline that's right on most parcels and confidently wrong on the rest is worse than one that flags the hard cases - because nobody double-checks the ones it got quietly wrong. You wouldn't be starting from a blank page. We have a live-tested approach to retrieving the underlying government spatial data, documented down to endpoints, parameters and a worked real-property example. That becomes your starting specification, along with sample property data to develop and validate against. THE CORE WORK - Constraint geometry: buildable envelope computation from parcel polygons, with setbacks applied per boundary type rather than as a uniform offset - Boundary classification - determining frontage, rear and side edges from geometry and supporting data - Coordinate reference system handling between source data, computation and web imagery - Robust behaviour on degenerate and invalid input geometry - Map composition and rendering: extent, scale and legibility at page size, with overlays correctly georeferenced onto retrieved imagery - A validation approach that demonstrates the output is right, not just that the code runs STACK (suggested) - Python, with Shapely - PostgreSQL + PostGIS - ArcGIS REST export, or a headless-browser rendering approach (Leaflet/MapLibre + Puppeteer) - open to either WHAT WE'RE LOOKING FOR A software engineer with deep geospatial expertise - someone who writes maintainable, tested code that runs unattended and produces output you'd stake your name on. Essential - Computational geometry on real spatial data - Shapely, GEOS, PostGIS or equivalent - Projections and coordinate reference systems, with an instinct for where they go wrong - Server-side map rendering, and the cartographic judgement to make output legible - Python, or a well-argued alternative - Clear written communication - you'll work directly with the technical lead, and documenting your reasoning is part of the deliverable Valuable - Messy real-world cadastral or parcel data, in any jurisdiction - Australian property, planning or proptech context - Sound judgement about uncertainty - several outcomes here are genuinely ambiguous, and surfacing that for human review beats emitting a confident-looking answer over shaky data Also of interest - mention it if any of this applies to you - Hands-on experience with commercial property-data or planning-assessment platforms, as a user or integrator rather than only as a developer, including a view on where their output is and isn't trustworthy - Building and deploying complete web applications end to end - API, database, deployment - if broader scope is something you'd want - Driving GIS tooling programmatically (PyQGIS, GDAL/OGR or similar) rather than only through a desktop interface Australian planning knowledge isn't required, that expertise sits on our side. HOW THE ENGAGEMENT WORKS Paid trial first. A representative slice of the real work: given a parcel polygon, attributes and imagery we provide, compute a setback envelope and render it to a single-page PDF. Up to 5 hours, paid at your own quoted rate - not a flat bounty, so it scales fairly wherever your rate sits. We're assessing thinking, validation approach and communication alongside geometry capability. Then the project itself. Currently scoped at around three months with a defined end point, rather than an open-ended retainer. If it goes well and further work comes up, we'd be glad to work together again. Rate is negotiable based on trial performance. We'd rather pay for capability than take the cheapest bid.