A mobile app and Laravel platform documenting damage with geotagged photos, using Gemini 3.5 Flash to normalize place names and classify damage, with an interactive map.
An Intelligent System for Supporting Reconstruction Decision-Making in Damaged Areas is an academic project carried out by a team of students with technical assistance from Techno Enjaz, to document damage in affected areas of Syria and analyze it with multimodal AI. Volunteers and field staff use a mobile app to collect reports containing geotagged photos, a text description, and the area name; a Laravel server then passes the reports to an AI Agent powered by the Gemini 3.5 Flash model, which normalizes area names and classifies the damage level from text and image together.
Processed results are shown in a web platform and dashboard with an interactive map of damage locations, statistics, and charts, so decision-makers can see how damage is distributed and how severe it is across districts and governorates and prioritize interventions. The system keeps raw field data exactly as received, separate from processed data, with a log of every AI request and response.
| Item | Details |
|---|---|
| Project type | Decision-support system for damage documentation: mobile app, web platform, and dashboard |
| Field | Multimodal AI, geographic information systems, humanitarian work and reconstruction |
| Project status | Implemented system with documented screens, tested with input scenarios simulating field work |
| Techno Enjaz role | Technical support and assistance to the students during project development |
| Core technologies | Flutter (Android), Laravel, MySQL, Queues, Gemini 3.5 Flash (text and vision), AI Agent, interactive maps |
| Outputs | Geotagged damage reports, normalized area names, damage level and percentage, interactive map, statistics and charts |
Years of conflict in Syria caused widespread destruction of infrastructure, housing, and services, with no unified databases. Most documentation relies on paper reports or unstructured field data gathered from multiple sources, and the project identifies its main problems:
The project was carried out academically by the students, with technical assistance and support from Techno Enjaz during development. It is presented among the office's work as an example of its support for systems that combine field data-collection apps, Laravel servers, multimodal AI models, and decision-support dashboards.
The agent is not a standalone model but a control layer between the server and the Gemini API: it prepares text and image inputs, sends them in a unified format, receives structured results, and enforces unified classification and assessment logic that limits variation between reports. The project does not develop AI models from scratch; it relies on ready-made APIs.
The agent analyzes the area name as entered by the user and compares it with standard names, so reports about the same location are merged into a single spatial entity. Examples shown in the project:
| User input | Normalized form | Governorate |
|---|---|---|
| ريف دمشق دوما (Rif Dimashq Douma) | ريف دمشق – دوما (Rif Dimashq – Douma) | Damascus |
| دوما الريف (Douma al-Rif) | ريف دمشق – دوما (Rif Dimashq – Douma) | Damascus |
| حلب السكري (Aleppo Sukkari) | محافظة حلب – حي السكري (Aleppo Governorate – Sukkari district) | Aleppo |
Linguistic cues of damage severity are extracted from the text description, and photos are analyzed with Gemini 3.5 Flash Vision to extract visual features such as collapses, cracks, and total destruction; the two sources are merged into a unified classification. When a sufficient text description is missing, the damage level can be estimated from the image alone. Each processed report stores the damage level, its percentage, the analysis description, and a confidence indicator.
| Table / entity | Role |
|---|---|
| Users | User data; no report exists without the user who created it |
| Raw reports | Data as entered: location name, coordinates, description, report status |
| Images | Multiple photos per report |
| AI log | Requests sent and responses received, for tracing and reviewing analysis |
| Processed reports | Damage level and percentage, analysis description, and confidence indicator, linked to a governorate and area |
| Governorates and areas | Reference tables for normalizing names and preventing duplicates |
| Part | Screens and functions |
|---|---|
| Website landing page | System concept in Arabic and English, features (AI damage analysis, GPS location tracking, photo and media upload), workflow steps, statistics |
| Admin screens | Dashboard with statistics and charts, latest reports, interactive map with pop-up info windows, report details and damage assessment, advanced search and filtering, adding, editing, and deleting reports, profile |
| User screens (web) | Tracking reports and their status, adding a report with map location, photos, PDF files, and video links, report details and analysis result, notifications, profile |
| Mobile app | Login, side menu, report lists in Arabic and English, filtering by status and damage level, three-step report creation, report details, drafts, profile editing |
The project documents a complete system made up of a mobile app for collecting field data and a central web platform for managing, analyzing, and displaying reports, with automated analysis of text and images and normalization of place names. Testing relied on input scenarios simulating field work, varying the quality of descriptions, place naming, and damage levels.
The project reports no quantitative measurements of damage-classification or name-normalization accuracy, processing time, or the number of reports tested. The results should therefore be read as a functional proof of an implemented system, not as an evaluation of classification accuracy under real operating conditions.
According to the project's recommendations:
The system collects photos and precise geographic locations of damaged areas, so any real deployment needs clear access controls, protection of photos and coordinates, and human review of AI classifications before they are used in planning and resource-allocation decisions.
If you are working on a decision-support system based on data and image analysis and need technical help developing the idea or building a prototype, you can contact the Techno Enjaz office to discuss your project's requirements and the appropriate scope of support.
It is an academic project built by a team of students with technical assistance from Techno Enjaz to document damage in affected areas of Syria and analyze it with multimodal AI. It includes a mobile app for field reports, a Laravel server with an AI agent based on Gemini 3.5 Flash, and a web platform and dashboard with an interactive map and statistics.
The Laravel server first stores the report as raw data without modification, then runs a background processing job via Queues so the user does not wait. The AI agent sends the text and images to Gemini 3.5 Flash, which normalizes the area name and classifies the damage level, and a processed report linked to a reference governorate and area is saved, or a failed status is recorded.
The AI agent compares the entered name with standard names; for example, 'Douma al-Rif' and 'Rif Dimashq Douma' both become 'Rif Dimashq – Douma'. Damage is classified by combining cues from the text description with visual features in the photos, such as collapses and cracks, and the damage level, percentage, analysis description, and a confidence indicator are stored.
The system uses a Flutter mobile app for Android, a Laravel server with a MySQL database and asynchronous processing via Queues, and the Gemini 3.5 Flash model for text and image analysis. The dashboard shows an interactive map of damage locations, statistics, and charts with advanced search and filtering.
The project documents an implemented system with a field app and a central web platform, tested with input scenarios that simulate field work, but it reports no quantitative measurements of damage-classification accuracy or processing time. Its limits include a focus on Syrian areas and Android, the need for an internet connection, and dependence on the external Gemini API.
Reach out to Techno Enjaz engineering team to discuss technical feasibility, prototype hardware, and development scope.
Discover more engineering prototypes and academic implementations
A children's learning system that uses a camera to recognize animals with YOLO11, colors with HSV, and numbers with OCR, then shows and speaks the result in Arabic.
A computer vision system that detects knives and pistols in real time from a laptop or phone camera using two trained YOLOv8 models, Python, and OpenCV.
An interactive learning system for children that recognizes objects via the camera with YOLOv8, speaks their names, detects colors with HSV, and labels uploaded images.
