Hamara HR Hygiene AI

Computer vision for deployed workforce

Built a sub-20MB computer-vision model that audits uniform and grooming standards across 5,000+ fuel stations and reports to a live dashboard.

01 — Project context

What exists today, and how this does it better.

FIG. 02

Overview

One of India's largest fuel retailers wanted brand consistency everywhere — uniform-hygiene standards held across thousands of forecourts, visible centrally and close to real time.

Manual audits couldn't do it. At national scale the travel and labour were simply too expensive to run often enough to matter.

The scale

5,000+
Retail stations to cover
5 lakh+
Attendants to audit

01 — The pipeline

Let the camera do the audit

I led the team that built the whole pipeline — computer vision to check uniform and grooming standards from a simple photo, so an audit takes a glance instead of a visit. We built a POC first to prove the results held up on real forecourts.

02 — The view

A city on one screen

The pilot shipped with a realtime dashboard — hygiene-audit scores across a whole city at a glance, flagged stations surfaced instantly. The same view scales cleanly from one city to the entire national network.

Built to scale cheaply

The model came in at under 20 MB — small enough to run almost anywhere, which is exactly what auditing five thousand stations affordably demands.

Five thousand stations, half a million people — audited from one screen.

My role

Led the team that built it end-to-end, the detection model, the realtime dashboard, and the pilot that proved a manual, country-wide process could become a glance.

Outcome

Content coming soon.

Stack

YOLO
Computer Vision
Edge (<20MB)
Realtime dashboard
PyTorch