Responsibilities
Model Development & Training
- Train and fine-tune computer vision models for use cases such as object detection, classification, segmentation, pose estimation, and video analytics
- Improve model performance through architecture tuning, optimisation strategies, and experimentation
- Work on data augmentation, loss tuning, learning rate scheduling, and training improvements
- Evaluate newer vision and multimodal AI approaches where applicable
Research & Experimentation
- Study and apply insights from technical papers and industry research
- Prototype and benchmark different modelling approaches against baseline systems
- Analyse experiment outcomes and document findings, limitations, and trade-offs
Model Evaluation & Debugging
- Investigate edge cases, prediction inconsistencies, and real-world model behaviour
- Troubleshoot issues such as false positives/negatives and unstable inference results
- Improve model robustness through better datasets, inference logic, and tuning strategies
Data & ML Infrastructure
- Work with large-scale datasets including preprocessing, validation, and quality checks
- Support experiment tracking, reproducibility, and dataset versioning workflows
- Develop maintainable experiment pipelines and configuration-based training setups
Deployment & MLOps
- Assist with deployment and monitoring of ML models in production
- Work with training, inference, and evaluation pipelines
- Support model lifecycle processes including retraining, validation, rollback, and version control
- Consider optimisation factors such as latency, GPU utilisation, and inference efficiency
Requirements
- Degree in Computer Science, AI, Machine Learning, or related field (or equivalent practical experience)
- Around 1–2+ years of hands-on experience in deep learning or computer vision projects
- Strong Python programming skills
- Experience with PyTorch and/or TensorFlow
- Familiarity with Linux environments and command-line usage
Nice-to-Have Skills
- Exposure to multimodal or open-vocabulary AI models
- Experience customising model architectures, layers, or loss functions
- Familiarity with tools such as Git, Docker, MLflow, or Weights & Biases
- Knowledge of ONNX, TensorRT, quantisation, or model optimisation techniques
- Experience handling video data using OpenCV or FFmpeg
