Working as a Data Scientist. I am a master's graduate in Biomedical Data Science from Imperial College London. I also have a Bachelor's in Electrical Engineering. I aspire to create ML applications and use my wide tech knowledge to create exciting things. You can look at some of my projects and interests on this page, and find my resume, and contact info.
A song recommendation tool which uses an indexing engine FAISS to create vector embeddings of song descriptions that match the vibe of a user. The training dataset was specifically created for the purpose of the project. It uses a Siamese transformer network to capture the sentiment of the songs.
This is a small Flask API application with one API serving a Pytorch model that calculates similarity between text and image captions. The models used are from the hugging face library and use the ViT-GPT2 transformer model for image captioning and Sentence transformers for encoding the text sentences. The image example is from the Flickr8k dataset. The two models are pre-trained and wrapped within a Pytorch module in order to enable training both models as a one Pytorch model. A training example for the sentence transformer is also given for fine-tuning.
A Python traning course for companies. Created as part of the Keitaro traning platform. The course involved video lectures and interactive code along jupyter notebook exercises.
The main aim was to create a robust, adaptable, and easy to use text-processing pipeline tool that would recognize assays in text, but also serve as a good starting point for the implementation of other algorithms and entity recognition tasks.
This project implements a pipeline of model agnostic explainability techniques both in 2D and 3D for medical images. The use-case data is a collection of CT chest scans as 3D nifti files from the MosMed COVID-19 publicly available dataset.
For further business inquries feel free to send me an email or contact me on social media