cv

Team Lead, AI

Hyperfine, CT, USA

• Pioneered a state-of-the-art deep learning (DL) MRI reconstruction algorithm and a novel training framework for reconstructing extremely noisy low-field MRI data with no ground truth.
• Delivered significant contributions to FDA 510(k) clearances for DL reconstruction and AI measurement tools for brain MRI, including designing and deploying models, developing training and evaluation schemes for noisy and heterogeneous dataset with limited labels, and devising and executing verification and validation protocols for the regulatory requirements.
• Conduct research and develop key innovations in model-based image reconstruction, unsupervised image denoising, reconstruction, and nonrigid registration, simulation-based model training frameworks.
• Mentor a team of senior scientists and interns to help them navigate the R&D process for projects including signal processing, sensor-data denoising, anatomical segmentation, and self-supervised learning

Staff Scientist

Hyperfine, CT, USA

Senior Scientist

Hyperfine, CT, USA

Intern

Hyperfine, CT, USA

Machine Learning Research Intern

Magic Pony, Twitter, London, UK

• Investigated learned index structure and approximate nearest neighbour systems to improve real-time content-based image retrieval system

Software Engineer Intern

Moore Europe Capital Management, London, UK

• Worked on front-end projects for their quasi real-time analytic infrastructures for financial analysis and econometrics. The technology involved JavaScript and React framework.

PhD, Computer Science

Imperial College London, UK

• Thesis - Deep Learning for Fast and Robust Medical Image Reconstruction and Analysis (link)
• Supervisors - Prof. Daniel Rueckert and Prof. Jo Hajnal.
• Specialisation - Deep Learning, Convolutional & Recurrent Neural Networks, Inverse Problems, Image Segmentation, Compressed Sensing, Magnetic Resonance Imaging.

MEng, Mathematics and Computer Science

Imperial College London, UK

• First Class Honours
• Thesis - Deep Belief Network A step towards modelling Attachment Theory

fastMRI Image Reconstruction Challenge

• 34 teams participated in the challenge of developing state-of-the-art MR image reconstruction techniques for large-scale knee MR dataset
• Placed 2nd, 3rd and 5th in “multicoil 4x”, “multicoil 8x” and “singlecoil 4x” tracks respectively. (paper)

STACOM Multi-sequence Cardiac MR Segmentation Challenge

• Placed 1st in the challenge of developing state-of-the-art techniques for segmenting myocardium provided limited data in multi-contrast. (paper)

• Dean's List - MEng. Mathematics and Computer Science (Yr2)