Computer Vision Engineer

Job Overview :

We are seeking a highly skilled Computer Vision Engineer specialising in real-time detection, pose estimation, and movement analysis to design and deliver robust vision-based intelligence systems. The role focuses on building scalable pipelines capable of detecting people and objects, analysing human movement patterns, and tracking individuals across frames in real-world environments. This position will play a critical role in developing production-grade systems that operate reliably across varying lighting conditions, camera angles, occlusions, and live video feeds, ensuring consistent performance in real deployment scenarios.

Key Responsibilities :

• Design and implement real-time computer vision pipelines for person and object detection.

• Develop advanced pose estimation and movement analysis models to interpret body posture, gestures, and motion patterns.

• Build multi-object tracking systems to accurately track individuals across frames and maintain identity persistence over time.

• Optimise models for low-latency inference suitable for edge and near-edge deployments.

• Ensure system robustness across diverse environmental conditions, including varying lighting, occlusions, and camera perspectives.

• Integrate computer vision models with video ingestion, streaming, and processing frameworks.

• Perform model evaluation and benchmarking using real-world datasets and live feeds.

• Collaborate with cross-functional teams to align vision models with business use cases and operational requirements.

• Continuously improve detection accuracy and tracking stability using state-of-the-art techniques.

Skills & Qualifications :

• Strong expertise in Computer Vision and Deep Learning techniques is required.

• The candidate should have hands-on experience with object detection approaches such as YOLO, Faster R-CNN, and SSD; pose estimation frameworks such as OpenPose, HRNet, and MediaPipe; and multi-object tracking methods such as DeepSORT, ByteTrack, and OC-SORT.

• The role requires proficiency in Python and deep learning frameworks such as PyTorch or TensorFlow, along with experience in real-time video processing pipelines using OpenCV, GStreamer, or FFmpeg.

• A strong understanding of image processing, feature extraction, model optimisation, GPU acceleration, and performance tuning is important for delivering reliable systems in production environments.

• Experience in edge deployment and inference optimisation using TensorRT, ONNX, and quantisation techniques will be valuable.

• The candidate should also be comfortable evaluating models using precision, recall, tracking accuracy, latency, and related metrics while balancing trade-offs between accuracy, performance, and cost.

Experience :

• 3 - 5 years of experience in computer vision engineering or applied AI development, with a proven track record of building and deploying real-time vision systems in production environments.

• The candidate should have practical experience working with live video feeds, real-world datasets, and deployment scenarios where lighting, camera angles, occlusions, and operating conditions may vary significantly.

What Success Looks Like :

• Success in this role will be measured by reliable detection and tracking performance across varying real-world conditions, high accuracy in pose and movement analysis with minimal drift, low-latency inference suitable for real-time applications, and stable system performance when tested against live feeds, lighting variations, and camera angles.

Future Focus & Impact :

In this role, you will contribute to building next-generation intelligent vision systems that go beyond detection to deliver context-aware insights from human movement and interactions. The platform will evolve to support higher levels of automation, behavioural analytics, and decision intelligence, enabling scalable deployment across environments with increasing complexity and volume. Your work will play a pivotal role in shaping a robust AI-driven foundation capable of handling real-world variability while maintaining precision, scalability, and operational efficiency.