I'm an accomplished and innovative Machine Learning Engineer with a passion for developing and delivering state-of-the-art deep learning models. I have a proven track record of success in training computer vision systems and deploying models on edge devices.
I have achieved significant milestones throughout my career, including:
- Delivering home surveillance object detection models for edge devices, enhancing security by accurately detecting people and animals in everyday settings.
- Training computer vision systems to detect quality issues on Jaguar Land Rover vehicles, resulting in a significant increase in accuracy compared to existing processes.
- Leading the development and delivery of lightweight Multi-Object Detector (MOD) networks with improved metrics like mean Average Precision and max F1 score.
- Driving the adoption of coding standards and implementing CI/CD practices to ensure code quality and reduce regression risks.
- Leading the consolidation of multiple network training repositories into a single python package, resulting in codebase size reduction and improved customer experience.
- Designed and trained generations of lightweight Multi-Object Detector (MOD) networks for home surveillance on edge devices, continuously improving key metrics using PyTorch and PyTorch Lightning framework.
- Conducted experiments to optimize network architecture, data augmentation, and optimization changes, integrating successful improvements into the MOD codebase.
- Spearheaded the establishment of team coding standards, including CI/CD stages for regular code linting and testing, ensuring better code quality and reducing regression risks.
- Led the delivery of training code as part of a Software Development Kit (SDK), streamlining the process for customers to train multiple networks using a single python package.
- Facilitated bi-weekly retrospectives following AGILE best practices, driving process improvements and fostering efficient experimentation within the team.
- Collaborated with cross-functional teams across different time zones to coordinate data acquisition initiatives, network quality assessments, and timely delivery to customers.
- Led the design of datasets to address network biases, resulting in improved performance in low-light scenarios and extended functionality for home security cameras.
- Machine Learning
- Deep Learning
- Computer Vision
- Python
- PyTorch
- TensorFlow
- Data Processing
- Model Deployment
You can connect with me on: