AI KnowledgeAdvancedEnded
DREAM-KG Homelessness Support
“Build dynamic knowledge graphs with explainable AI to connect homeless individuals with real-time resources.”
The Challenge
The challenge is to explore the KG to understand data gaps and enhancements the community can bring to the KG, so that we end up with challenge ideas that other resource groups can then address.
Timeline
DREAM - Challenge Kickoff & KG Exploration
KickoffFri, Apr 3, 2026
Official launch of the DREAM-KG Homelessness Support challenge with introduction to the knowledge graph and data exploration objectives.
DREAM - Data Gap Analysis Milestone
MilestoneTue, Apr 7, 2026
Teams submit initial findings on data gaps and potential enhancements identified in the knowledge graph.
DREAM - Mentor Session: Challenge Ideation
Mentor SessionTue, Apr 14, 2026
Expert mentors provide guidance on refining challenge ideas and ensuring they address key community needs for homelessness support.
DREAM - Final Solution Posts
DeadlineFri, Apr 17, 2026
Teams submit their final challenge solutions!
About This Challenge
The Homelessness Crisis: A Data and Coordination Challenge Every night, hundreds of thousands of Americans experience homelessness, while service providers, social workers, and first responders struggle with fragmented, outdated information about available resources. Shelter beds go unfilled while people sleep outside just blocks away. Critical services like mental health support, job training, and transitional housing operate in information silos. The lack of real-time, coordinated data creates a cascade of missed opportunities that perpetuates the cycle of homelessness and wastes precious resources. The human cost is very high, individually and socially. Individuals and families unable to access help that exists but remains invisible to them and their advocates. The economic impact is equally staggering, with cities spending billions on emergency interventions that could be prevented through better coordination and resource allocation. The Technical Challenge: Building Adaptive Knowledge Systems Current homeless service databases are static, siloed, and quickly become obsolete. Shelter capacity changes hourly. Programs open and close. Eligibility requirements shift. Contact information becomes invalid. What's needed is a Dynamic, Responsive, Adaptive, and Multifaceted Knowledge Graph (DREAM-KG) that can continuously learn, update, and maintain accurate information across multiple sources while providing explainable AI recommendations. Your mission is to develop intelligent model editing approaches that can: - Reconcile conflicting information from multiple datasets and knowledge sources - Automatically detect and correct outdated information while preserving historical context - Learn and integrate new knowledge about services, locations, and pathways to care - Maintain data integrity while enabling real-time updates from cities and nonprofits - Provide explainable recommendations for optimal service pathways Impact and Implementation This is a master-level challenge designed as a 2-4 week intensive project that can spawn multiple focused mini-challenges. Your work will directly contribute to the NSF Proto-OKN DREAM-KG initiative, with real potential for deployment in cities across multiple states. The ultimate goal is to create a living knowledge system that cities and nonprofits can contribute to and benefit from - turning fragmented data into coordinated action. Field workers will access this through mobile interfaces to find real-time bed availability, social workers will identify comprehensive service pathways, and researchers will gain unprecedented insights into effective intervention strategies. Your Deliverable Teams will produce an enhanced, adaptive knowledge graph system with demonstrated capabilities for multi-source data integration, intelligent error correction, and real-time updates. The best solutions will be integrated into ongoing research partnerships with cities and shelter organizations, with potential for scaling across the national Continuum of Care network.
Research Team
This challenge emerges from the NSF Proto Open Knowledge Network (Proto-OKN) DREAM-KG initiative, a cutting-edge research program focused on developing dynamic knowledge graph systems for homelessness service coordination. The research is led by Dr. Yuzhou Chen at the University of Memphis, whose expertise in knowledge graph systems, dynamic data integration, and AI-driven information reconciliation directly addresses the complex challenges of maintaining accurate, real-time service information in rapidly changing social service environments. Principal Investigator: Dr. Yuzhou Chen, Assistant Professor of Computer Science at the University of Memphis, brings specialized expertise in knowledge graph construction, multi-source data fusion, and explainable AI systems. His research portfolio includes groundbreaking work on adaptive knowledge systems that can maintain accuracy and relevance in dynamic environments - precisely the technical challenge faced in homelessness service coordination where information changes hourly and consistency across multiple data sources is critical. Proto-OKN Program Context: The National Science Foundation's Proto Open Knowledge Network program specifically funds projects that develop interconnected, continuously-updated knowledge networks serving the public good. Dr. Chen's DREAM-KG initiative represents a flagship application of this vision, targeting one of society's most pressing coordination challenges through advanced AI techniques. Research Innovation Areas: Dr. Chen's team is pioneering several breakthrough approaches directly applicable to this challenge: - Dynamic Knowledge Graph Maintenance: Developing systems that automatically detect inconsistencies in shelter capacity, service availability, and program eligibility without requiring manual oversight - Multi-Source Data Reconciliation: Creating algorithms that intelligently merge conflicting information from city databases, nonprofit records, and real-time updates while maintaining complete data provenance - Contextual Model Editing: Building systems that can update knowledge graphs while preserving critical historical relationships and service pathway dependencies - Explainable AI for Social Services: Ensuring that service recommendations can be clearly articulated to social workers, case managers, and clients Collaborative Research Network: The Proto-OKN initiative emphasizes real-world impact through multi-stakeholder collaboration. Dr. Chen's research directly involves municipal data teams, homeless service providers, and social work practitioners, ensuring that technical innovations address genuine operational needs in the field. This challenge offers participants the unique opportunity to contribute to NSF-funded research that bridges advanced AI techniques with urgent humanitarian needs, working alongside established investigators to create systems with immediate real-world deployment potential.
Resources
# Data Sources for DREAM-KG Homelessness Support Challenge ## Federal Government Datasets ### HUD (Housing and Urban Development) - HUD Point-in-Time Count Data - Annual homeless population estimates by geography - URL: https://www.hudexchange.info/resource/3031/pit-and-hic-data-since-2007/ - Continuum of Care (CoC) Program Data - Geographic boundaries and funding information - URL: https://www.hudexchange.info/programs/coc/ - Housing Inventory Count (HIC) Data - Bed capacity and utilization by program type - URL: https://www.hudexchange.info/resource/4832/2022-hic-data-by-coc/ - HUD Exchange Resource Database - Federal programs and resources - URL: https://www.hudexchange.info/resources/ ### Other Federal Sources - SAMHSA Treatment Locator API - Mental health and substance abuse treatment facilities - URL: https://findtreatment.gov/ - VA Homeless Programs Directory - Veterans-specific services and housing - URL: https://www.va.gov/homeless/ - USICH Federal Strategic Plan Data - Policy frameworks and coordination efforts - URL: https://www.usich.gov/federal-strategic-plan/ ## Open Data Portals (City/County Level) ### Major Cities with Comprehensive Homeless Data - Los Angeles Homeless Management Information System (HMIS) - URL: https://data.lacity.org/ (search "homeless") - San Francisco DataSF - Homelessness - URL: https://data.sfgov.org/browse?category=Housing+and+Buildings - Seattle Open Data - Human Services - URL: https://data.seattle.gov/browse?category=Community - New York City Open Data - DHS Services - URL: https://opendata.cityofnewyork.us/ - Chicago Data Portal - Human Services - URL: https://data.cityofchicago.org/browse?category=Health+%26+Human+Services ### Regional Data Sources - Bay Area Regional Data Commons - Multi-county homelessness data - URL: https://www.bayareadata.org/ - King County (Seattle) Homelessness Data - URL: https://kingcounty.gov/depts/community-human-services/COVID/sheltering-plan/data.aspx ## APIs and Real-Time Data Sources ### Service Directories - 211 API - Comprehensive social services database (varies by region) - URL: https://211.org/ (contact local 211 for API access) - iCarol Platform API - Powers many 211 implementations - URL: https://www.icarol.com/ - United Way API - Local resource databases - URL: https://www.unitedway.org/ (regional APIs available) ### Geospatial and Infrastructure - Google Places API - Location verification and business hours - URL: https://developers.google.com/maps/documentation/places/web-service - OpenStreetMap Overpass API - Geographic data for services and amenities - URL: https://overpass-api.de/ - Census Bureau APIs - Demographics and geographic boundaries - URL: https://www.census.gov/data/developers/data-sets.html ## Research Repositories and Academic Sources ### Homelessness Research - National Alliance to End Homelessness Data Hub - URL: https://endhomelessness.org/homelessness-in-america/homelessness-statistics/ - Chapin Hall at University of Chicago - Longitudinal homelessness studies - URL: https://www.chapinhall.org/research/homelessness/ - USC Center for Homelessness, Housing and Health Equity - URL: https://uschhhe.org/ ### Knowledge Graph and Ontology Resources - Schema.org Social Services Vocabulary - URL: https://schema.org/ - Open Eligibility Standard - Standardized eligibility criteria - URL: https://openeligibility.org/ - Human Services Data Specification (HSDS) - URL: https://docs.openreferral.org/en/latest/hsds/ ## Specialized Databases ### Healthcare and Mental Health - HRSA Data Warehouse - Federally Qualified Health Centers (FQHCs) - URL: https://data.hrsa.gov/ - NAMI Resource Directory - Mental health services - URL: https://www.nami.org/find-support - SAMHSA Behavioral Health Treatment Locator - URL: https://findtreatment.gov/ ### Legal and Benefits - LawHelp.org Database - Legal aid services for low-income individuals - URL: https://www.lawhelp.org/ - Benefits.gov API - Federal benefit program eligibility - URL: https://www.benefits.gov/ - National Immigration Forum Resources - Immigration-related services - URL: https://immigrationforum.org/ ## Non-Profit and Community Data ### National Organizations - National Coalition for the Homeless Directory - URL: https://nationalhomeless.org/references/need-help/ - Salvation Army Locations API - URL: Contact: https://www.salvationarmyusa.org/usn/ - Catholic Charities USA Directory - URL: https://www.catholiccharitiesusa.org/ ### Specialized Populations - National Safe Place Network - Youth services - URL: https://www.nationalsafeplace.org/ - National Network for Youth Directory - URL: https://www.nn4youth.org/ - Women's Shelters Directory - URL: https://www.womenshelters.org/ ## Research and Survey Data ### Academic Research Datasets - ICPSR Homelessness Studies - University of Michigan data archive - URL: https://www.icpsr.umich.edu/web/pages/ - **Urban
Skills
You'll Learn
Systems ThinkingPythonData AnalysisKnowledge ExtractionSocial ImpactData VisualizationKnowledge Graph DesignBusiness AnalysisGraphQL for Knowledge GraphsCritical ThinkingPredictive Analytics
Helpful to Have
Neo4j
Hosted by
Mahesh
0 teams / 50 max
Starts Apr 3, 2026 (EDT)
Ends Apr 19, 2026, 11:59 PM (EDT)