Profile
Available

Smart Queue Management System

Category
Applied AI / Full-Stack
Year
2025
Role
Personal Project — Applied AI
Context
Personal project — applied AI exploration
Smart Queue Management System

Tools & Technologies

DjangoReactPostgreSQLClaude API (LLM Integration)REST APIs

Approach & Engineering

I built a full-stack queue management system with a Django REST backend, a PostgreSQL database modeling queues, counters and visitor flow, and a React front end for both staff and visitors. On top of the core system, I integrated the Claude API so staff can ask natural-language questions about the state of the queue ("which counter is falling behind?", "what's our average wait time this hour?") and get direct, data-grounded answers instead of manually filtering dashboards.

Vision and Impact

Most queue systems show raw numbers — tickets waiting, counters open — but leave the interpretation to the human. I wanted to build a system where the data and an LLM work together: the backend owns the ground truth, and the LLM turns that truth into plain-language insight and concrete recommendations for rebalancing flow.

Key Technical Challenges

The central challenge was grounding: making sure the LLM only reasons over real, current data from PostgreSQL rather than hallucinating queue statistics. I designed a query layer that fetches precise, scoped data (per counter, per time window) and passes it to the Claude API as structured context, so natural-language answers stay accurate even as the queue state changes in real time.

Solving Complex Problems

I had to solve real-time synchronization between the Django backend and the React front end so queue status stays current for both visitors and staff, design a data model that generalizes across different queue types (single-line, multi-counter), and keep LLM calls fast and cost-effective by only sending the minimal relevant context rather than the entire dataset.

User-Centric Design

Two very different users share this system: visitors, who need a simple, low-friction way to see their position and estimated wait time, and staff, who need fast operational insight. The interface is deliberately split — a minimal visitor view and a richer staff view with the natural-language assistant.

Meeting Stakeholder Needs

Visitors needed transparency about wait times without asking staff. Staff needed a way to spot bottlenecks and rebalance counters without manually cross-referencing numbers. The Claude integration closes that gap: staff can ask a direct question and get a decision-ready answer grounded in the live queue data.