Work

Aptiv Interior Sensing Portfolio

Deep Learning
Computer Vision
Automotive

Major contributor to Aptiv's interior sensing portfolio, including gesture recognition systems deployed in BMW and driver monitoring technology.

Pioneering In-Cabin AI

As a Deep Learning Engineer at Aptiv, I was a major contributor to the company’s interior sensing portfolio—a comprehensive camera-based platform that interprets driver behavior, gestures, and attention state.

Gesture Recognition System

Aptiv developed the technology behind the first gesture recognition system for automotive applications, introduced in the BMW 7 Series in 2015.

How It Works

  1. 3D Zone Capture: Camera mounted in roof module captures hand positions and motions
  2. Infrared Illumination: LEDs provide clear imaging in all lighting conditions
  3. Real-Time Analysis: Computer vision and machine learning translate gestures to commands
  4. Command Library: Predetermined gesture library for intuitive control

Supported Gestures

  • Single finger rotation for volume/zoom control
  • Pointing gestures to accept calls
  • Swipe gestures to reject calls
  • Two-finger gestures for audio control
  • Pinch gestures for display manipulation

Driver Monitoring System

The interior sensing platform also powers driver monitoring capabilities, enabling:

  • Driver Recognition: Automatic identification and preference adjustment
  • Attention Monitoring: Real-time assessment of driver focus
  • Mood Detection: Behavioral analysis for adaptive vehicle response
  • Emergency Intervention: Automatic braking in critical situations
  • Autonomous Handoff: Seamless transition between autonomous and manual driving

OEM Deployments

This technology has been deployed across multiple major automotive manufacturers:

  • BMW: Gesture recognition in 7 Series and subsequent models
  • GM: Driver Attention Assist in Cadillac Escalade IQ, Chevrolet Tahoe
  • Multiple Tier 1 suppliers worldwide

Technical Innovation

  • Developed deep learning models for robust hand tracking
  • Optimized inference pipelines for automotive-grade embedded systems
  • Created training data pipelines processing millions of in-cabin images
  • Achieved real-time performance on resource-constrained hardware