
CROSSCHECK: Using Amazon Bedrock to Detect Public Infrastructure Collisions
CROSSCHECK is an infrastructure intelligence platform that uses Amazon Bedrock to extract project data from government work-order documents and detect potential conflicts between road, drainage, water, electricity, and telecom projects.
CROSSCHECK: Making Public Infrastructure Projects Work Together
CROSSCHECK is an infrastructure intelligence platform that detects potential collisions between public infrastructure projects and helps people understand the infrastructure activity behind everyday traffic disruptions.
When youβre stuck in traffic, you often know there is congestion ahead, but not why. It could be road resurfacing, drainage work, water pipelines, utility maintenance, or multiple projects happening on the same road. Similarly, when a navigation app asks you to change your route, it usually tells you that a faster route exists, but doesnβt clearly explain what is causing the disruption ahead.
CROSSCHECK connects these missing pieces.
Government agencies often plan road resurfacing, drainage, water pipelines, electricity, and telecom work independently, with information spread across tenders and work-order PDFs. CROSSCHECK uses AI to extract project details such as agency, work type, location, dates, and cost, then combines spatial and temporal analysis to identify projects that may affect the same road during overlapping time periods.
The platform visualizes these projects on an interactive map, helping users understand what is happening on the road ahead, why a particular area may be congested, and which infrastructure projects could be contributing to the disruption.
For infrastructure planners and government agencies, CROSSCHECK provides visibility into potential project conflicts before construction begins. For everyday commuters, it provides context behind road disruptions instead of simply saying βtraffic aheadβ or βrerouting.β
The Problem
Public infrastructure projects often involve multiple agencies working independently.
A single road can have projects related to:
Road resurfacing
Drainage
Water pipelines
Electricity
Telecom infrastructure
Road widening
Utility maintenance
When project information is distributed across different documents and departments, it becomes difficult to identify conflicts before work begins.
This can result in:
Repeated road excavation
Construction delays
Increased public disruption
Rework
Wasted expenditure
Poor coordination between agencies
The key question we wanted to answer was:
What if we could automatically detect these potential collisions before construction starts?
Introducing CROSSCHECK
CROSSCHECK turns unstructured infrastructure documents into structured, searchable project intelligence.
The platform follows a simple pipeline:
Government Documents β AI Extraction β Project Data β Spatial & Temporal Analysis β Collision Detection β Interactive Map
A user can upload a work-order or tender document, and CROSSCHECK extracts important information such as:
Agency
Work type
Road or location
Project dates
Project cost
Other relevant project details
The extracted information is then compared with existing projects to identify potential conflicts.
Using Amazon Bedrock for Document Intelligence
One of the biggest challenges was converting unstructured government documents into structured information.
Government work orders are not necessarily formatted like databases. Important information can appear inside paragraphs, tables, or different document layouts.
We use Amazon Bedrock to process these documents and extract structured project information.
For example, a document might contain a location such as:
Cross Road No. 4, near Ayyappa Temple, Ward 76
Instead of treating this as plain text, CROSSCHECK transforms the information into structured project data that can be used for further analysis.
This allows the rest of the system to work with infrastructure projects as data rather than simply as documents.
Spatial + Temporal Collision Detection
Extracting the information is only the first step.
The more interesting problem is determining whether two projects actually have the potential to interfere with each other.
CROSSCHECK evaluates projects using two dimensions:
- Spatial overlap Are the projects happening on the same road or nearby road segments?
- Temporal overlap Are the projects scheduled during overlapping or closely related time periods? A project that is happening on the same road but several years apart may not represent the same risk as two projects scheduled for the same month. By combining spatial and temporal information, CROSSCHECK can identify potential collision scenarios. The Collision Map We wanted the output to be more intuitive than a spreadsheet full of project records. CROSSCHECK provides an interactive map where infrastructure projects can be visualized geographically. Projects are categorized based on their collision risk, allowing users to quickly identify areas that may require coordination. The map helps answer questions such as: Which roads have multiple upcoming projects? Which projects overlap geographically? Which projects have overlapping schedules? Where could construction disruption be concentrated? How much project expenditure could potentially be affected? Instead of manually comparing hundreds of documents, users can visually explore potential conflicts. Estimated Expenditure at Risk Another important part of CROSSCHECK is understanding the potential financial impact. When multiple projects collide, the issue isn't only construction inconvenience. There can also be additional costs associated with: Rework Repeated excavation Delays Rescheduling Resource utilization CROSSCHECK therefore calculates an estimated expenditure associated with projects involved in potential collisions. This provides another way to prioritize which conflicts may require attention. Explainability Matters Infrastructure decisions require more than a red marker on a map. We therefore designed CROSSCHECK with source evidence in mind. When a potential collision is identified, users can trace the project information back to the source document that was used to extract it. This creates a simple chain: Source Document β Extracted Data β Collision Analysis β Map Result The goal is to make the system easier to inspect and validate rather than treating the AI output as a black box. AWS Architecture The core workflow uses AWS services to process and analyze infrastructure documents. At a high level:
- The prototype uses several AWS components across the workflow.
βββββββββββββββββββββββ
β Government Documentsβ
ββββββββββββ¬βββββββββββ
β
ββββββββββββββββ
β Amazon S3 β
ββββββββ¬ββββββββ
β
ββββββββββββββββββββββ
β Amazon Bedrock β
β Document Extractionβ
ββββββββββββ¬ββββββββββ
β
ββββββββββββββββββββββ
β Structured Project β
β Data β
ββββββββββββ¬ββββββββββ
β
ββββββββββββββββ
β DynamoDB β
ββββββββ¬ββββββββ
β
βββββββββββββββββββββββββββ
β Spatial + Temporal β
β Collision Analysis β
βββββββββββββ¬ββββββββββββββ
β
ββββββββββββββββββββ
β Interactive Map β
ββββββββββββββββββββThe architecture is intentionally modular.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article