Kavach: An AI System That Watches for the Moment a Driver's Attention Starts to Slip
Nayara Energy moves fuel to over 6,300 retail outlets using more than 10,000 tankers, and driver fatigue and rash driving put lives at risk every day. ibs Fulcro built Kavach, an AI and computer vision system that reads a driver's face in real time, warns them the moment attention drifts, and alerts a live control room before a warning sign becomes an accident.
At a glance
Client: Nayara Energy (Oil & Gas / Fleet Safety)
Problem: Nayara Energy's fleet of 10,000+ fuel tankers faced real, daily accident risk from driver fatigue and rash driving, with no real-time way to detect and intervene.
Solution: Kavach: a real AI and computer vision system reading driver attention in real time, alerting both the driver and a live control room before a warning sign becomes an accident.
Result: Reduction in road accidents due to human error: 85%
ibs Fulcro's role: AI & Computer Vision
The Problem
Nayara Energy is on a mission to keep India on the move, transporting petroleum products to more than 6,300 retail outlets nationwide using a fleet of over 10,000 tankers. But driver fatigue, drowsiness, and rash driving are frequent, serious issues on India's roads, resulting in accidents that endanger drivers' lives every single day.
Nayara Energy wanted to use technology to meaningfully improve fleet driver safety, not by reacting after an incident, but by predicting and preventing it before it happened.
The Solution: Kavach
ibs Fulcro built Kavach, an AI and computer vision driver protection system. At its core is an AI and ML-powered sensor that continuously monitors driver behaviour and attention, built around four connected functions: continuous monitoring, analysis, real-time warnings, and human intervention.
By analysing facial patterns, eyes, head gaze, and other natural cues, Kavach distinguishes between safe, attentive driving and the early signs of drowsiness or inattention, well before those signs become a crash.
The technical rigor behind “analysing facial patterns”: Kavach's underlying model performs genuine Facial Action Unit coding, quantifying specific, named facial movements, brow raises, cheek raises, lid tightening, jaw drop, and others, each scored with a real confidence percentage, mapped against a facial landmark mesh tracked in real time. This isn't a simple eyes-open-or-closed check; it's a structured behavioural signal built from dozens of individually measured facial cues.
Beyond Drowsiness: A Broader Behaviour Model
Kavach doesn't only watch for fatigue. The system is trained to identify three additional unsafe behaviours directly linked to distracted driving: smoking, drinking, and phone use, each detected from the same continuous camera feed used for fatigue monitoring, giving Nayara's safety team a single, unified view of driver risk rather than separate systems for separate behaviours.
What Happens the Moment Risk Is Detected
When Kavach detects unsafe behaviour, it issues real-time acoustic and visual alerts directly to the driver, aiming to interrupt the risk before it escalates into a road accident. When severe fatigue specifically is detected, Kavach goes a step further, connecting directly to Nayara's centralised fleet monitoring control room. From there, human operators communicate with the driver via built-in two-way radios, guiding them to the nearest Nayara pump for rest and recuperation.
Every unsafe event Kavach detects is also logged as data, feeding continuous AI-based insights that refine the behaviour monitoring engine itself over time, so the system improves with every tanker on the road rather than staying static after deployment.
How Kavach Works, End to End
| Stage | What Happens |
|---|---|
| Continuous monitoring | An AI and ML-powered in-cabin sensor tracks facial patterns, eye behaviour, and head gaze in real time |
| Analysis | Facial Action Unit coding and a tracked facial landmark mesh distinguish safe driving from drowsiness, distraction, smoking, drinking, or phone use |
| Real-time warnings | Acoustic and visual alerts are issued directly to the driver the moment unsafe behaviour is detected |
| Human intervention | Severe fatigue triggers a direct connection to Nayara's centralised control room, where operators guide the driver to the nearest pump via two-way radio |
| Continuous learning | Every unsafe event is logged and fed back into the system, refining the behaviour monitoring engine over time |
Source: campaign video, ibs Fulcro / Nayara Energy.
The Results
Kavach has been installed in 5,000 tankers, roughly half of Nayara's total fleet, and has analysed over 2 million kilometres of driving data. The impact has been substantial: an 85% reduction in road accidents due to human error, a 30% reduction in risky driving behaviour overall, and more than 1,000 severe accidents averted.
Why this is the number that matters most, beyond the percentages. An 85% reduction rate is a strong statistic, but it's abstract. Averting more than 1,000 severe accidents is a concrete count of real incidents, each one involving a real driver's life, that didn't happen because the system caught a warning sign in time. For a fleet safety product, that's the actual outcome everything else in the system exists to produce.
Why This Matters Beyond One Fleet
Kavach serves as trusted protection that operates 24/7 to keep Nayara's partners safe and their families secure. The broader lesson for any operator running a large, geographically dispersed fleet is that meaningful safety improvement doesn't require replacing drivers with automation, it requires giving human drivers a genuinely attentive second set of eyes on their own state, backed by a real human control room ready to intervene the moment automated detection isn't enough on its own.
Frequently asked questions
Kavach is an AI and computer vision driver protection system built by ibs Fulcro for Nayara Energy, using an in-cabin sensor to continuously monitor fleet drivers for fatigue, drowsiness, distraction, smoking, drinking, and phone use, issuing real-time alerts and connecting severe cases directly to a centralised control room.
Kavach analyses facial patterns including eyes, head gaze, and other natural cues, using Facial Action Unit coding, quantifying specific facial movements such as brow raises, cheek raises, and jaw drop, mapped against a real-time facial landmark mesh, to distinguish safe driving from signs of drowsiness or inattention.
Yes. Beyond fatigue and inattention, Kavach is trained to identify smoking, drinking, and phone use from the same continuous camera feed, giving a unified view of driver risk rather than separate detection systems.
Kavach issues real-time acoustic and visual alerts to the driver directly, and for severe fatigue specifically, connects to Nayara's centralised fleet monitoring control room, where human operators communicate with the driver via built-in two-way radios and guide them to the nearest Nayara pump to rest.
Kavach has been installed in 5,000 tankers and analysed over 2 million kilometres of driving data, delivering an 85% reduction in road accidents due to human error, a 30% reduction in risky driving behaviour, and more than 1,000 severe accidents averted.
Nayara Energy transports petroleum products to more than 6,300 retail outlets nationwide using a fleet of over 10,000 tankers. Kavach has currently been enabled across 5,000 of those tankers, roughly half the fleet.