Physics-Aware AI Surveillance System

Physics-Aware AI for Video Surveillance

Project Overview
We are looking for an experienced AI/Computer Vision developer or researcher to develop a prototype of a Physics-Aware AI-based Video Surveillance System. The objective is to build a surveillance solution that combines deep learning, computer vision, object tracking, motion analysis, and physics-based constraints to detect and predict abnormal events more accurately than conventional AI systems.

Client Requirements
The system should: Detect and track people, vehicles, and other objects in CCTV/video footage.
Analyze position, velocity, acceleration, direction, and trajectories.

Detect abnormal activities such as:
Falls
Unauthorized entry
Abnormal movement
Vehicle–pedestrian collision risks
Unusual trajectories
Crowd-related incidents
Predict potential incidents before they occur, where possible.
Incorporate physics/kinematic constraints to identify physically implausible or anomalous motion.
Generate explainable alerts, e.g., why an event was classified as high risk.
Provide confidence/risk scores for detected events.
Include basic privacy-preserving mechanisms, such as face/person anonymization where required.
Provide a visualization/dashboard showing detected objects, trajectories, and alerts.

Expected Technical Skills : Python, PyTorch, Deep learning, Physical AI

Preferred Candidate
Candidates with a background in Computer Vision, Physics-Informed Machine Learning, Video Analytics, AI/ML research, or Intelligent Surveillance Systems will be preferred.
Please include relevant previous projects/GitHub/demo links and briefly explain your experience with physics-aware AI or video analytics.

For background on our company (KAAI-VISIONS), feel free to browse the current projects here: https://sites.google.com/view/kaai-vision917/home.

If you’d like updates on future consultancy openings, US based projects or part-time roles, subscribe to our channel: https://youtube.com/@kaai-visions?si=xMNj_MugNtN-Su1a.