Results-oriented AI Consultant with extensive experience in machine learning, computer vision, and data science, specialising in advanced algorithms, model optimisation, and high-performance computing. Proven ability to enhance particle detection in cryoEM samples and develop MLOps systems that drive significant revenue growth and operational efficiency. Skilled in deploying NLP-based solutions, including sentiment analysis and bias detection, with a strong background in consulting and developing innovative AI frameworks. Passionate robotics instructor and published researcher with a solid academic foundation in Data Science, adept at utilising diverse tools and frameworks to deliver impactful data-driven insights.
– Explored cutting-edge object detection algorithms (DynamicDet, YOLOv7, UNet)
and applied denoising techniques (Topaz, filtering, contrast enhancement) to
improve particle picking in cryoEM protein samples.
– Conducted ablation studies and optimised particle detection, improving model precision
by 0.3 and recall by 0.25.
– Utilised high-performance computing clusters with Bash scripting and command-line
deployment to enhance computer vision applications in PyTorch.
– Passionate robotics instructor teaching Arduino coding and fostering logical thinking in
young learners.
– Developed the backend code structure for PACE-ML, including modules for data processing,
training pipelines, performance and fairness metrics, incremental learning,
hyperparameter tuning, and workflow orchestration using Airflow.
– Created machine learning pipelines for home loan application processing, integrating
Responsible and Fair AI solutions that improved pipeline efficiency and increased
revenue by $150k.
– Consulted on and developed an Aspect-Based Sentiment Analysis pipeline for AWS
Marketplace Listings, creating over 20 listings and achieving a 35% increase in
efficiency.
– Contributed to the Organisations Quantum Challenge, developing a Trainable Quantum
Machine Learning algorithm for image classification, which enhanced accuracy by 7%.