Full-Stack / GIS Engineer for Geospatial AI Dashboard

Full-Stack / GIS Engineer for Geospatial AI Dashboard We are developing a geospatial inspection platform for municipalities and public-space asset management. Street-level imagery is collected from a moving vehicle and processed using computer vision to identify roadside objects and classify vegetation. The results are then presented in an interactive, map-based dashboard. The data collection and AI processing for our first municipal survey have already been completed. We are now looking for an experienced Full-Stack / GIS Engineer to take the processed output and integrate it properly into our existing dashboard. This is a clearly defined first project within a larger development roadmap. We want to start with this specific integration and, if the collaboration works well, potentially continue with the engineer on the broader platform. The Assignment A survey has already been processed and contains: * 514 georeferenced image frames; * approximately 3,700 raw object detections; * clustered detection data; * vegetation classifications and segments; * GPS information per frame; * detection classes and confidence scores; * JSON and GeoJSON outputs; * vegetation masks; * dashboard-specific exports and conversion scripts. You do not need to develop or retrain the AI models. Your work starts with the processed output. We already have an existing web dashboard with an interactive map. Your task is to make this dataset work properly within that existing application. The main objective is to create a clear connection between: Map → location → detection → image → classification & details A municipal user should be able to open the survey, navigate the map, understand what was detected and inspect the corresponding imagery without needing any technical knowledge of the underlying AI. What we need you to do You will first review the existing codebase, dashboard and supplied dataset to understand the current architecture and identify what needs to be improved. You will then: * integrate the processed survey into the existing application; * correctly display the surveyed route, object detections and vegetation information; * implement sensible clustering/zoom behaviour so thousands of detections do not overwhelm the map; * connect detections and map locations to the corresponding captured imagery; * provide useful filtering by object/vegetation class, confidence and other relevant attributes; * create a clear detail view for individual detections and frames; * improve the relevant map interactions, information panels and general UX/UI; * ensure acceptable performance with the complete dataset; * keep the implementation maintainable and reusable for future survey datasets. We are not looking for a complete redesign or rebuild. The objective is to improve and complete this specific part of the existing application. One important data consideration is that the current GPS coordinates represent the vehicle/camera position at the moment of capture, rather than the exact physical position of each detected object. The interface should preserve this distinction rather than presenting the detections as exact surveyed object coordinates. Expected Result At the end of the assignment, we should be able to open the municipal survey in the existing dashboard and professionally explore the results. The map should be clean and intuitive, detections and vegetation should be displayed correctly, filters should work, and users should be able to move naturally between the geographical information and the underlying survey imagery. The implementation should also establish a sensible structure for adding future surveys without creating municipality-specific hard-coded solutions. We expect clean, maintainable code and a short technical handover explaining the implemented data flow and how another survey following the same structure can be added. Who We Are Looking For We are looking for an engineer, not a pure UI designer or machine-learning specialist. Relevant experience includes: * full-stack web development; * React / JavaScript / TypeScript; * GIS and web mapping; * Mapbox, MapLibre, Leaflet or similar; * GeoJSON and spatial data; * APIs and backend integration; * PostgreSQL/PostGIS where relevant; * data visualization; * image-based inspection applications; * integration of computer-vision outputs. Experience connecting map features, imagery and AI-generated detections is particularly valuable. For the first milestone, we would like you to run the existing application, inspect the supplied dataset and current data flow, identify the required changes and propose a practical implementation approach before making major changes. The Bigger Picture This assignment is intentionally a self-contained development chunk, but it sits within a larger product roadmap. Our longer-term goal is a scalable platform where municipal survey runs can move through a consistent workflow: Data collection → AI processing → geospatial processing → dashboard → inspection & validation → usable municipal asset information Future development could include automated ingestion of new surveys, multiple municipalities and projects, comparison between survey rounds, change detection, human validation of AI detections, more accurate object positioning, LiDAR integration, asset management, reporting, municipal GIS integrations, APIs and scalable cloud infrastructure. We therefore want to solve the current integration properly first, while building it in a way that does not block this future development. If this first project goes well, there is potential for substantial follow-up work. When Applying Please briefly tell us about your experience with GIS/map-based web applications, GeoJSON and interactive mapping. We would particularly like to see 1–3 examples where you have worked with large numbers of map features, connected map data to imagery, or integrated computer-vision/data-processing outputs into a web application. Also briefly explain how you would approach an existing application with several thousand detections that need to remain fast, clear and easy to explore. Please start your proposal with “GEOSPATIAL” so we know you have read the complete description.