Emmanuel Moses Mpyambala
AI & Data Science UDOMAutoPresence AI : Workplace Analytics System
Problem Statement
Current government and institutional attendance systems rely on manual logbooks or standalone biometric solutions such as fingerprint scanners and sometimes calling names manually, all these are inefficient and unreliable at scale / enterprise scale operations.. These systems suffer from physical wear and failure rates, slow throughput during peak entry hours which might lead to serious wastage of potential time More critically, traditional systems only capture static entry timestamps, providing no visibility into whether employees remain on duty throughout working hours. This creates opportunities for absenteeism, payroll fraud, and ghost attendance, where recorded presence does not reflect actual workplace activity. Additionally, biometric systems introduce long term privacy risks due to the centralized storage of irreversible identity data.
Proposed Solution
AutoPresence AI solves this by automatically recording employee presence using existing office cameras. It runs in the background, detects when staff arrive and leave, tracks how long individuals are actually present in the office and generates accurate attendance records without requiring any manual check in. Most importantly, it’s built with privacy in mind so no video is stored or viewed, only attendance analytics data is recorded. The result is faster entry, zero queues, reduced time fraud, and a smarter, more transparent way to manage workforce attendance and overall workplace employees performance analytics and predictions
Innovation
1.No New Hardware Requirement - Utilizes existing CCTV infrastructure via RTSP/ONVIF streams, eliminating the need for dedicated biometric devices. 2.High Concurrency Detection - Supports multi-person tracking in dense environments without requiring scanning or physical interaction. 3.Edge-Local Privacy Architecture - All processing is performed on local compute nodes. Raw video frames are never stored, streamed, or viewed. Instead, frames are immediately converted into non-invertible feature embeddings and discarded in memory and only structured event metadata is persisted. 4.Beyond Check-In Systems - Moves beyond binary attendance by continuously analyzing presence duration, movement patterns, and duty-zone occupancy for more accurate workforce accountability analytics and insights. 5.Data Minimization by Design - No raw biometric templates or video recordings are retained, significantly reducing privacy and breach risk compared to fingerprint systems. 6.On - device computer vision for real-time detection
Expected Impact
1.no more fingerprint queues, filling books and no more manual calling of names 2.No more fake attendance records 3.Clearer understanding of who is actually present at work and their performance 4.Detecting Intruders 5.Faster and smoother office entry process 6.Saves time and increase efficiency on work places 7. Unannounced absenteeism alerts,
Standing
Team Members
Mentor
Lusajo Menaro Shitindi