I investigate diffusion-based methods for generating synthetic medical imaging data in low-data regimes. As part of this work, I introduced DermaFlux, a framework that improves pathology classification performance by 8% through the curation of a 500k image–text pair dataset. In addition, I explore the use of large language models (LLMs) to generate synthetic medical captions and automate the classification of pathology reports.

As a Computer Vision Engineer at Huawei UK, I specialized in 3D facial reconstruction from monocular images. In this role, I integrated cutting-edge techniques to advance the field, focusing on enhancing the accuracy and effectiveness of reconstruction methods. My work involved pushing the boundaries of state-of-the-art approaches, such as NeRF and diffusion-based techniques. Moreover, I focused on designing a generic 3D Gaussian Splatting framework for animatable head avatars, enabling consistent identity generation across viewpoints.

I worked as a Research Assistan(RA) in the project ARISE held by Business School, Imperial College of London. ARISE was a European Union-funded initiative designed to forecast agricultural crop yields within a specific region during targeted time periods. My responsibilities encompassed the utilization of data derived from satellites and weather stations, employing state-of-the-art machine learning algorithms to extract meaningful insights. Additionally, I was tasked with generating synthetic data for regions with limited data availability, ensuring a comprehensive and robust approach to yield prediction.

My main responsibilities included the design and implementation of cutting-edge automated pipelines for collecting images across the web. These pipelines were instrumental in the creation of novel datasets that accurately represented real-world scenarios. In addition, I took charge of designing and coordinating human annotation tasks for the annotators at ArielAI while ensuring precise and consistent annotations..

In my capacity as the sole Machine Learning Engineer, I undertook the development of a robust network for automated product recognition within images captured from supermarket shelves. This required designing Computer Vision algorithms and tools for easy annotation and creating both training and detection procedures alongside with back-end support.