Built on a government-grade agentic architecture, SafeRoads AI brings together multimodal data and AI-assisted decision support to help government agencies make more informed road safety decisions.
Led by
Johns Hopkins University + Partners
Built for
Municipal transportation agencies
Focus
Road safety decision support
SafeRoads AI
From fragmented municipal data to decision-ready road safety intelligence
Project activity
The SafeRoads AI team, together with our government partner from the City of Bellevue, attended the official kickoff for the Google.org Impact Challenge: AI for Government Innovation. The week marked the launch of the Accelerator Program, bringing participating teams together with Google engineers and product experts to advance practical, human-centered AI solutions for public service.
Read Google's announcementPlatform architecture
A city-operated Data Hub provides the authoritative analytical foundation. The Agentic AI Layer orchestrates validated tools, authoritative evidence retrieval, Safe System filtering, and human-reviewed decision support.
Cross-cutting safeguards
Human decision authority · data minimization · privacy and security · reproducible methods · accountable deployment
Use cases
Illustrative applications
Data fusion
Harmonize crash, roadway, mobility, speed, enforcement, project-history, video-derived indicators, Google Earth AI and other geospatial and contextual data in a canonical spatial-temporal schema.
Risk characterization
Use validated SQL, GIS, statistical, and contextual evidence to support network screening, corridor evaluation, and transparent prioritization.
Decision support
Retrieve authoritative engineering evidence, assess local applicability and Safe System alignment, and support human-reviewed decisions and performance monitoring.
Evidence + resources
Government-approved road safety guidance and evidence used to ground countermeasure analysis, including FHWA resources, CMF evidence, and HSM methods.
Validation studies and applied research supporting the platform's analytical and evaluation methods.
Architecture, data standards, analytical methods, Safe System logic, governance, and implementation guidance.
Open-source code, templates, evaluation harnesses, and deployment materials released as development progresses.
Team
Project leadership
Project partners
Municipal partners · Year 1
Work with us
