London Area, United Kingdom
For the better part of a decade, I have worked on AI research for both vision and natural language. During my PhD at University of Cambridge, I worked on deep learning-based super-resolution, image segmentation and image classification, and I have developed a number of AI methods for optical microscopy. In Oxford, I worked with the biotechnology company ONI, where I continued to develop computational methods for imaging using AI. I'm currently in AI Research at GSK working on large language models and agentic workflows. Other interests include mathematical modelling, artificial intelligence and deep learning.
Research on drug discovery using LLMs and agentic workflows. Multi-agent reasoning and orchestration. Contributed to GSK's GenAI system, Jules. Accelerated multi-node inference for in-house pre-trained LLMs, focusing on parallelisation and constrained decoding for agentic use.
Co-founded the non-profit organisation, Open-seneca, with the aim of mapping air quality across the globe. Started as a research project during our master's degree programme and spun out in 2019 with five fellow PhD students.
Developed and deployed deep learning models for spectral demixing, image segmentation, and super-resolution towards new product offerings. This involved R&D and MLOps. In addition to this, I have also contributed to a web-based acquisition and analysis software which involves several technologies spanning front-end, back-end and deployment.
I co-founded Wizion AI with a colleague from my research group, a company offering image processing solutions utilising deep learning methods inspired by research in our lab. In 2020, we were selected for the Cambridge Accelerate accelerator programme at the Judge Business School. After three years of efforts towards the development of two products, we decided to shift our focus to other opportunities.
Imaging at high spatio-temporal resolution requires a trade-off with image quality leading to low signal-to-noise ratio in acquired data. This renders traditional image analysis methods to perform unreliably. In my PhD thesis, I proposed methods for image reconstruction, denoising and segmentation using deep learning methods that are robust to noise.