NeighborNet

Inspiration

Every day, usable food and resources are wasted while people nearby struggle to access them.

A restaurant may have surplus meals at the end of the day. A family may have extra food after an event. A company may have unused supplies. At the same time, another household or community may urgently need those resources.

The problem is not always a lack of resources — it is the lack of trusted coordination between people who have something to give and people who need it.

We built NeighborNet to solve this coordination gap through AI-assisted orchestration, location-based matching, verification, and traceable delivery.

NeighborNet is designed for everyday community resource sharing, while also providing Disaster Mode for urgent community response during emergencies.

What it does

NeighborNet is an AI-powered community resource-sharing platform that connects surplus resources with verified community needs.

Restaurants, households, companies, and event organizers can donate surplus food and other resources. Verified recipients can request support, while nearby volunteers can transport donations when needed.

NeighborNet creates a trusted chain:

Verified Request → Donation → Volunteer Matching → Pickup Verification → Delivery → Recipient Confirmation

Every donation receives a unique Donation ID, allowing it to be tracked throughout the workflow.

Everyday Community Mode

NeighborNet coordinates normal community resource sharing by:

Matching available donations with verified requests
Finding suitable nearby volunteers
Considering location, capacity, urgency, and availability
Preventing over-allocation
Reserving tasks for the first valid volunteer
Verifying pickup and delivery using OTPs
Tracking task progress
Recovering and reassigning tasks when volunteers become unavailable
Maintaining an auditable history of important actions

Disaster Mode

When an emergency occurs, NeighborNet can switch to Disaster Mode.

The platform can coordinate:

Urgent community requests
Disaster-affected zones
Available resources
Volunteer dispatch
Task assignment and reassignment
Risk classification
Delivery verification
Operational status tracking
Human/admin oversight

The same infrastructure can therefore support both everyday community assistance and emergency response.

How we built it

NeighborNet is a full-stack platform with an AI agent orchestration layer.

Core technologies

Amazon Bedrock
Strands Agents SDK
FastAPI
Next.js
Location-based matching
OTP verification
Persistent data support
Audit logging

The most important architectural decision was separating AI orchestration from critical operational decisions.

The AI agent determines which tools and workflows should be used, while deterministic components handle decisions that must be predictable, safe, and auditable.

Deterministic operational components

VolunteerMatcher Finds suitable volunteers based on operational constraints such as location, availability, capacity, and task requirements.

PlanningEngine Creates and coordinates delivery tasks.

RecoveryEngine Handles failed, cancelled, or unavailable volunteer assignments and supports reassignment.

RiskClassifier Classifies operational situations into:

GREEN → AMBER → RED

This architecture gives us the flexibility of AI without allowing an LLM to independently make critical allocation or safety decisions.

Trust & Verification

Trust is a core part of NeighborNet.

A simple Delivered status is not enough to prove that a donation actually reached the intended recipient.

NeighborNet therefore uses a verification chain based on:

Verified community requests
Unique Donation IDs
OTP-based pickup verification
OTP-based delivery verification
Controlled task state transitions
Volunteer assignment ownership
Audit trails
Recovery and reassignment logic
Human/admin oversight

This creates a traceable chain from the original request to the final delivery.

Challenges we ran into

1. How much autonomy should the AI have?

Our biggest challenge was deciding what the AI should control.

Allowing an AI model to independently decide who receives resources or which volunteer should handle a task could produce unpredictable and difficult-to-audit decisions.

We solved this by separating:

AI → orchestration and workflow decisions

from

Deterministic software → operational decisions

and

Humans → important oversight and intervention

This allowed us to use AI where it provides value while keeping critical operations predictable.

2. Building trust

Another major challenge was proving that a resource actually moved through the system as intended.

We addressed this with verified requests, unique Donation IDs, OTP verification, controlled task states, and audit logs.

3. Supporting both normal and emergency situations

Community assistance and disaster response have different urgency and coordination requirements.

Instead of building two separate systems, we designed NeighborNet so the same core infrastructure can support both everyday resource sharing and emergency response.

Accomplishments we're proud of

We are proud of building an end-to-end workflow rather than just an AI chatbot.

Key accomplishments include:

Built an end-to-end surplus resource sharing workflow
Created location-based volunteer matching
Implemented verified community requests
Designed a request → donation → pickup → delivery verification chain
Added unique Donation IDs for traceability
Implemented first-valid-volunteer task reservation to prevent duplicate assignments
Added OTP verification for pickup and delivery
Implemented controlled task state transitions
Added automatic recovery and reassignment
Created GREEN / AMBER / RED risk classification
Added real-time task status tracking
Reused the same infrastructure for everyday support and disaster response
Added audit trails for accountability
Designed AI orchestration around deterministic operational safeguards

Our biggest accomplishment is turning a simple idea —

"Someone has extra. Someone nearby needs it."

— into a coordinated and trustworthy system.

What we learned

We learned that building an effective AI agent system is not simply about connecting an LLM to tools.

The more important question is:

What should the AI decide, what should software decide, and when should a human decide?

We learned that deterministic logic is essential for operations such as:

Matching
Capacity checks
Task assignment
Verification
Recovery
Risk handling

We also learned that trust must be designed into the workflow, rather than added later.

Verification, traceability, controlled status transitions, and human oversight make an AI-powered system significantly more reliable.

What's next

We plan to expand NeighborNet with:

Smarter demand and surplus prediction
Food expiry and freshness tracking
Multi-donation route optimization
Community trust and reputation scoring
Advanced anomaly and fraud detection
Offline support for low-connectivity areas
Improved emergency coordination
NGO and community organization integrations
Restaurant and business integrations
Impact analytics showing resources recovered and communities supported

Our long-term goal is to make NeighborNet a trusted community coordination infrastructure where surplus resources can be connected to genuine needs quickly, safely, and transparently — both in everyday life and during emergencies.