USED BY 50+ EMPLOYEES AT BHAGWAN PARSHURAM INSTITUTE OF TECHNOLOGY
BRACTANCE – Employee Territory Range Attendance Application
An AI-driven employee attendance app with geofence security and a facial recognition tracker.
The Overview
The Problem: Most organisations lack secure, automated mechanisms for validating presence, resulting in fraudulent check-ins and unreliable data that limits workforce analytics and decision-making. Moreover, most workplaces rely on fragmented systems like separate apps for attendance, leave tracking, reports, and communication, creating friction, inconsistency, and room for misuse.
The Solution: The Employee Attendance Tracker is a unified, AI-powered platform designed to automate attendance, location validation, and workforce insights using computer vision and intelligent language models. The system was built to eliminate manual logging and reduce fraud while simplifying everyday tasks with built-in reporting and chatbots.
Workflow of the application
Tech Stack
Key Features
🤖 Face Recognition Pipeline
- Built with CNN-based embeddings and clustering techniques for robust face verification.
- Optimised inference for real-time validation during sign-in/sign-out events.
📍 Geofencing & Integrity Validation
- Used device GPS to enforce location-bound attendance; it is only accepted when users are physically present within the organisation’s coordinates.
- Implemented continuous spoofing checks to block VPN-based or off-site punch-ins.
📂 Natural Language Assistance
Integrated a Seq2Seq / BERT-based chatbot to help employees with queries about leave balance, monthly attendance, policy FAQs, and more.
Use Case Diagram
Platform Capabilities
- Automated time logging with identity and timestamp captured via live camera.
- Geofencing, which allows marking attendance in a confined area. Includes security features to restrict false location.
- User dashboard integrated with AI chatbot for queries. Annual Reports & Insights using different graphical representations of user data.
Login Page
Technical & Research Takeaways
- Deploying computer vision models outside controlled settings revealed challenges, including varying lighting environments and false positives in look-alike scenarios.
- Explored model calibration, embedding distances, and clustering thresholds.
- Sparked curiosity in few-shot learning, feature robustness, and privacy-preserving vision systems.