The Challenge
Managing multiple Paying Guest (PG) facilities is a logistical nightmare involving manual attendance tracking, disjointed billing, and massive food wastage due to inaccurate meal planning. Existing software lacked advanced automation, multi-tenant support for a single owner, and intelligent forecasting.
PGPulse was built to solve this by consolidating all operations—inventory, subscriptions, residents, and menus—under one unified dashboard, enhanced with machine learning models that automatically track attendance and predict exact food requirements.
Engineering & Architecture
PGPulse was designed from the ground up for scalability, utilizing a containerized microservices architecture. The system effortlessly manages real-time state via React Query and Zustand while the Django backend orchestrates complex RBAC and AI inference pipelines.
System Architecture
Key Features
Face Recognition Attendance
Utilizing InsightFace to extract 512-dimensional facial embeddings, residents can securely mark entry and exit by scanning their faces via the frontend webcam interface.
AI Meal Prediction
A Scikit-Learn Random Forest classifier analyzes historical attendance, weather patterns, and the day of the week to predict exact meal counts, drastically reducing food waste.
Multi-Tenant RBAC
Complete Role-Based Access Control allowing owners to effortlessly switch contexts between multiple PG branches without needing separate logins.
Smart Inventory Tracking
Centralized tracking of groceries and supplies, with automated critical alerts triggered when stock levels dip below dynamically configured minimum thresholds.