Machine Learning Engineer (Python) (Remote), Cape Town

ENVIRONMENT:

A Machine Learning Engineer with a strong foundation in computer vision is sought by a provider of cutting-edge Tech Applications. Your core role will be to improve the quality and reliability of the deployed detection and classification models. This is a hands-on role focused on model refinement, error analysis, and data-driven performance improvements rather than building from scratch. The successful incumbent will require 2+ years of Python & Machine Learning experience including computer vision Deep Learning models. You also need to be familiar with model serving and inference pipelines (e.g., NVIDIA Triton Inference Server, ONNX, TensorRT).

DUTIES:

• Analyse and improve the performance of existing object detection and image classification models deployed in production.
• Systematically investigate missed detections and false alarms across diverse CCTV environments, identify failure patterns, and propose targeted fixes.
• Design and implement data augmentation strategies tailored to real-world CCTV challenges such as varying lighting, camera angles, resolution, weather conditions, and occlusion.
• Run controlled experiments to evaluate the impact of training strategies, hyperparameter changes, data balancing, and architectural tweaks on model performance.
• Contribute to the development and refinement of false positive filtering pipelines, including ensemble and verification-based approaches.
• Assist with data labelling workflows, quality checks, and dataset preparation for training and evaluation.
• Maintain clear records of experiments, results, and model performance metrics to support reproducibility and team knowledge sharing.
• Perform research into latest AI/ML techniques that bring business value.
• Work on ML Backend development for the production system and support infrastructure.

REQUIREMENTS:

• 2+ Years of experience with Python programming and general Machine Learning.
• Experience with computer vision Deep Learning models.
• Familiarity with model serving and inference pipelines (e.g., NVIDIA Triton Inference Server, ONNX, TensorRT).
• Exposure to MLOps tools such as Weights & Biases, MLflow, or similar for experiment tracking.
• Familiarity with annotation tools and labelling workflows (e.g., CVAT, Label Studio).