CV

Education

DPhil in Computer Vision and Machine Learning, University of Oxford, 2022 (Thesis)
B.S. in Computer Science, Mathematics, and Statistics, Maynooth University, 2017

Developed computer vision solutions for autonomous shopping problems including:

Camera Pose Estimation: developed camera pose estimation to enable rapid deployment of new facilities, built 3D environment reconstructions, and automatic camera selection to highlight human activity and product visibility.
2D/3D Object and Person Detection: from concept to production deployment. Including dataset construction and management, training pipelines, experiment management, efficient implementations and inference optimisation, deployment in a Kubernetes environment.
3D from images: developed systems to improve autonomous understanding of indoor environments to enable improved activity recognition, significantly reduce deployment time, while increasing accuracy, and enhanced client technical demonstrations.
Led a team working on 3D geometry, mentored junior members, collaborated deeply with other teams, formulated research problems, designed and conducted experiments, gave company wide presentations of the latest research products.

Research on Generative AI for mathematical reasoning, evaluating reasoning in accuracy sensitive domains:

Large Language and Vision Language Models: evaluating performance of existing models in specific domains, training and fine tuning to improve performance even with minimal data, defining research directions
Technical guidance: efficient computational resource utilisation, model optimisation, distributed training, mentoring students with diverse backgrounds on model architectures, and programming techniques

Developed and maintained data pipelines, data analysis, and generated insights for business customers:

Built pipelines to collect, clean, and store data from multiple sources
Developed and maintained data analysis tools and dashboards
Generated insights for business, and presented findings to non-technical audiences
Gernerated reports for business on customer behaviour, WiFi usage, useful insights for customer retention

Programming: Python, Java, Bash (highly experienced); Swift, Objective-C, HTML/Javascript/CSS, GoLang.
Frameworks: PyTorch, PyTorch-Lightning, NumPy, SciPy, Open3D, OR-Tools, MLX, Tensorflow, Pandas, REST API (Flask).
Tools: Git, Docker, Linux, Kubernetes, Azure, Jira, Confluence, CI/CD, AWS, Databricks.

Y Yang, A Bean, R McCraith, A Mahdi. Instruction tuning for large language models: the impact of human-inspired learning strategies. To be submitted to COLM (2024).
K Korgul, A M. Bean, F Krones, R McCraith, A Mahdi, Exploring The landscape of Large Language Models In Medical Question Answering: Observations and Open Questions. arXiv:2402.02460 (2024), submitted to NEJM AI.
R McCraith, E Insafutdinov, L Neumann, A Vedaldi, Lifting 2D Object Locations to 3D by Discounting LiDAR Outliers across Objects and Views. ICRA (2022).
R McCraith, E Insafutdinov, L Neumann, A Vedaldi, Direct LiDAR-based object detector training from automated 2D detections. NeurIPS Machine Learning for Autonomous Driving workshop (2023).
R McCraith, L Neumann, A Vedaldi, Calibrating Self-supervised Monocular Depth Estimation. NeurIPS Machine Learning for Autonomous Driving workshop (2022).
R McCraith, L Neumann, A Vedaldi, Real Time Monocular Vehicle Velocity Estimation using Synthetic Data. IEEE Intelligent Vehicles (2021).
R McCraith, L Neumann, A Zisserman, A Vedaldi, Monocular Depth Estimation with Self-supervised Instance Adaption. arXiv:2004.05821 (2020).

MLX Image Models: implementations of popular image models in MLX library
Annotator: Small Python application enabling annotation for semantic segmentation
LaText: iMessage App that allowed sending, receiving and editing LaTex formatted equations
Cox Box: Swift iPhone app which read from GPS, accelerometer, and bluetooth heart rate monitor to create records of rowing workouts with a live statistics and post activity summary with CoreData and iCloud synchronisation
Swift Machine Learning: Swift implementation of some machine learning algorithms, with SIMD optimisation
Stocks Mac: Mac Application to load and plot stock prices and basic analysis (Objective-C)

Computer Vision and Machine Learning, Engineering Science, University of Oxford (2019,2020,2021). Teaching Assistant, offered to DPhil (PhD) students in computer science and engineering science as part of EPSRC Centre for Doctoral Training in Autonomous Intelligent Machines and Systems.
Artificial Intelligence, Computer Science Department, University of Oxford (2019, 2020, 2021). Teaching Assistant, offered to undergraduates and MSc students in computer science.
Functional Programming for undergraduates, Computer Science Department, University of Oxford (2019, 2020, 2021). Teaching Assistant, offered to undergraduates in computer science.
Data Structures and Algorithms, Computer Science Department, University of Oxford (2019, 2020, 2021). Teaching Assistant, offered to undergraduates in computer science.
Advanced Language Modelling Methods, Oxford Internet Institute, University of Oxford (October 2023 - February 2024). Organised and delivered hands-on tutorial on training and fine-tuning language models.

College Sports Representative, Kellogg College, University of Oxford (2018-2019)
Christ Church College Rowing Club President, University of Oxford (2018-2019)
Mentored kids at Coder dojo (beginner programming club), 2011-2017