Team Nexus

An AI system that connects construction projects, vendors, and resources to predict contention, explain risks, and recommend actions before delays happen.

Comment

Inspiration

Construction companies rarely manage a single project. They manage portfolios of dozens of projects, all competing for the same vendors, materials, equipment, and logistics resources.

We noticed a hidden problem: most construction platforms help teams answer "What is happening in my project?" but not "How is another project's decision affecting mine?"

A rush order on one site can silently consume a vendor's capacity and disrupt another project's schedule days later. The delay is only visible after the damage is already done.

This inspired us to build NEXUS — an AI-powered portfolio intelligence layer that sees the connections between projects and detects conflicts before they become delays.

What it does

NEXUS builds a shared resource intelligence graph across multiple construction projects.

It connects:

  • Projects
  • Vendors
  • Materials
  • Crews
  • Logistics routes
  • Purchase commitments

Using this graph, NEXUS detects hidden resource contention events where multiple projects compete for limited capacity.

Instead of only predicting a delay, NEXUS identifies the root cause:

"Project B's new steel order exceeded Vendor X's capacity, putting Project A's delivery timeline at risk."

The system then:

  1. Detects cross-project resource conflicts
  2. Traces the causal impact chain
  3. Explains the risk in natural language
  4. Generates mitigation options
  5. Allows a human decision-maker to approve the recommended action

NEXUS transforms construction supply chains from reactive tracking systems into proactive decision systems.

How we built it

We designed NEXUS as a portfolio-level AI decision system with multiple layers:

1. Data Intelligence Layer

We created synthetic construction portfolio data representing:

  • Multiple active projects
  • Shared vendors
  • Material orders
  • Delivery schedules
  • Resource capacities

The prototype simulates the type of information available from procurement and project management systems.

2. Shared Resource Graph

We modeled the construction ecosystem as a graph:

  • Nodes represent projects, vendors, materials, crews, and routes
  • Edges represent dependencies, commitments, and constraints

This graph allows NEXUS to reason across project boundaries instead of viewing each project independently.

3. Contention Detection Engine

The system continuously evaluates demand versus available capacity.

For a resource (R):

$$ Contention\ Risk = \frac{Total\ Demand}{Available\ Capacity} $$

When overlapping commitments exceed capacity, NEXUS identifies the conflict and evaluates downstream impact.

4. AI Explanation and Decision Layer

An LLM-powered agent converts graph insights into human-readable explanations and drafts mitigation actions such as:

  • Reallocating surplus material
  • Activating alternate vendors
  • Adjusting project schedules

Every action includes a human approval step to maintain trust and control.

5. Visualization Dashboard

We built an interactive interface that shows:

  • Portfolio health
  • Resource dependencies
  • Risk propagation
  • Recommended actions

The dashboard demonstrates how a single decision can create a ripple effect across projects.

Challenges we ran into

Modeling a problem without a public dataset

There is no publicly available dataset that represents real construction portfolio-level resource contention. We addressed this by creating realistic synthetic scenarios based on common construction supply chain patterns.

Balancing AI complexity with hackathon feasibility

A production system would require ERP integrations, real-time data pipelines, and extensive historical data. For the prototype, we focused on demonstrating the core intelligence: detecting hidden dependencies and explaining their impact.

Making AI decisions trustworthy

Construction decisions involve significant cost and safety implications. Instead of creating a fully autonomous system, we designed NEXUS with a human-in-the-loop approach where AI recommends and humans approve.

Accomplishments that we're proud of

  • Built a new way of looking at construction supply chains: from isolated projects to a connected portfolio system
  • Created a working graph-based representation of shared construction resources
  • Demonstrated detection of hidden cross-project conflicts before visible delays occur
  • Combined graph reasoning, forecasting concepts, and agentic AI into one decision-support workflow
  • Designed an explainable AI system rather than a black-box prediction model

The biggest accomplishment is proving that construction delays can be addressed earlier by finding the hidden conflicts that create them.

What we learned

We learned that the hardest problems in enterprise AI are not always about generating answers — they are about understanding relationships.

A single project may appear healthy in isolation while being at risk because of decisions happening elsewhere.

Building NEXUS taught us the importance of:

  • Modeling real-world dependencies
  • Designing explainable AI systems
  • Combining traditional algorithms with generative AI
  • Building solutions around human workflows, not just technology

What's next for NEXUS

Our vision is to evolve NEXUS into a real-time intelligence layer for large construction portfolios.

Future enhancements include:

  • Live ERP and procurement integrations
  • Advanced graph-based risk prediction models
  • Real-time vendor capacity monitoring
  • AI-driven optimization of material allocation
  • Multi-project simulation and "what-if" planning

NEXUS aims to help construction companies move from discovering problems after they happen to preventing them before they begin.

Built With

Share this project:

Updates

Submission history