At a glance
What was built, on what, and why it matters.
- Industry
- Construction
- Domain
- Infrastructure Project Delivery
- Solution
- Knowledge Graph Intelligence · Conversational AI · Project Intelligence
- Technology partner
- AWS
- AI accelerator
- Amazon Bedrock Agent
Key outcomes
Measured where it counts.
- 0124× faster graph build
- 0225,000 project activities ingested and queried
- 03Query responses in seconds, with source citations
- 04Delivered in 8 weeks
Critical data, fragmented.
Large infrastructure projects generate vast amounts of data across schedules, risk registers, contracts, and documentation — critical to decisions, yet scattered across systems.
Answering questions about how risks, activities, and contractual obligations relate meant manual investigation across sources, taking anywhere from 30 minutes to several hours. An earlier attempt had failed to scale, creating the need for an architecture that could operate across a complete project dataset.
One connected knowledge graph.
Graph-first architecture on Neo4j + Amazon Bedrock.
AWS co-funded a proof of concept and selected DBiz as delivery partner to validate whether a knowledge-graph-powered platform could operate at full infrastructure-project scale.
Built graph-first on Neo4j and AWS, the solution unified schedules, risk registers, work breakdown structures, and contract documentation into one connected graph. Amazon Bedrock provided the LLM capabilities for entity extraction, relationship inference, and conversational access.
- Automated ingestion of schedules, risk registers, and documentation
- Cross-source linking via deterministic rules + AI semantic inference
- Conversational querying with source-cited responses
- Delay-propagation and risk-impact analysis across connected data
- Native vector indexing for semantic search and retrieval
Connected insight, in seconds.
The engagement validated knowledge-graph technology for large-scale infrastructure intelligence. Teams now query connected insight across schedules, risks, contracts, and documentation through a single model.
It established a scalable foundation for broader deployment, and demonstrated extensions like voice interaction and live transcript ingestion for richer, field-based project intelligence.
Technology stack
| Layer | Technology |
|---|---|
| Cloud platform | AWS |
| AI & generative AI | Amazon Bedrock · Amazon Bedrock Agent |
| Data processing | AWS Lambda · AWS Fargate · AWS Step Functions |
| Graph database | Neo4j (native vector indexing) |
| Core capabilities | Entity extraction · Relationship inference · Conversational retrieval |