Applied ML Engineer- Accident Detection (G-Sensor + Video AI)

Job Specification:

Applied ML Engineer — Accident Detection (Dashcams)

Company: Kasava AI Private Limited

Location: Office-Based — Kolkata, India

Reports to: India Tech Lead / CTO

Role Type: Full-time

1. Role Overview

Kasava builds AI-powered dashcams and fleet safety systems for business customers across the UK. We have over 10,000+ connected dashcams deployed and customers are actively demanding automatic accident detection and incident reporting.

Our devices already capture:

• Video clips
• GPS and speed metadata
• X/Y/Z accelerometer (G-sensor) impact readings

We are hiring an Applied ML Engineer to own and deliver accident detection models that work reliably in real fleet environments.

This role is focused on shipping production ML, not research.

2. Role Mission

Your mission is to build an accident detection system that:

 detects true collisions and serious incidents

 avoids false positives (bumps, potholes, harsh braking)

 generates actionable alerts + video evidence for fleet managers

improves continuously with real-world data

3. Key Responsibilities

Accident Detection Model Development

• Build classification models using accelerometer X/Y/Z data
• Detect impact signatures and collision patterns
• Develop probability scoring: accident vs non-accident

Sensor + Event Feature Engineering

• Engineer features such as:
• spike magnitude (√x²+y²+z²)
• jerk (rate of change)
• multi-axis impact patterns
• sustained abnormal events (rollover/tilt)
• Reduce false positives from road noise

Data Pipeline & Labeling Support

• Work with engineering to build accident-event datasets
• Define labeling strategy for triggered clips
• Help bootstrap training data from fleet events

Deployment & Production Integration

• Collaborate with backend engineers to integrate ML into the platform
• Deliver real-time or near-real-time incident alerts
• Monitor model performance and drift over time

Future Expansion (Phase 2)

• Extend sensor-based detection into video confirmation:
• crash visual cues
• stopped vehicle patterns
• scene anomaly detection

4. Required Qualifications

Must Have

• 3–7 years experience in applied machine learning
• Strong experience working with time-series or sensor data
• Feature engineering expertise (accelerometers/IMU data preferred)
• Proficiency in Python and ML frameworks:
• PyTorch or TensorFlow
• 
  
• Ability to build practical models that perform in messy real-world environments
• Strong communication and ownership mindset
• Must be office-based in Kolkata

5. Preferred / Bonus Skills

• Computer vision / video classification experience
• Experience with IoT, automotive, telematics, surveillance devices
• Familiarity with deployment workflows (Docker, APIs, edge constraints)
• Knowledge of anomaly detection and event-based ML systems
• Exposure to MLOps practices (monitoring, retraining pipelines)