Developing a novel methodology to make pain quantifiable and objective
Encoding data in high-dimensional latent space and employing Topological Data Analysis (TDA) on manifolds
20 million people worldwide suffer from drug-resistant epilepsy. For severe cases, the only remedy is removing where the seizures tend to originate from (the Epileptogenic Zone) with surgery, hence (as an extension of a research project) designed a:
Physics-Informed Neural Network that predicts where the Epileptogenic Zone is, locally in your browser
Integrated the 4 Hodgkin-Huxley equations (that describe the properties of the axon membrane) into the model's loss function to abide by brain thermodynamics
Lets you interact with a 3D WebGL model of the brain (and predicted seizure origin) with gestures
Click here
Engineered an interactive Model UN study quiz to master flag knowledge and the official names of UN Member States (not 'Bolivia', but 'Plurinational State of Bolivia')
Ideation -> Deployment in 2 days
Click here
Web app which teachers at my school used to create subject-specific content such as worksheets and practice questions
Utilised the Google AI Studio API to provide responses
Constructed 5 prompt engineering courses, guiding students on how to effectively prompt AI
3,500 word essay explaining how bias in AI arises and deliberating its potential impact on society in the next few years, and in the coming decades
Awarded a Distinction in my internal school competition
Click here
The brain is the most complex object in the known universe - it's fundamentally a biological supercomputer - unravelling the brain's complexity represents an elusive challenge for humanity
Researching machine learning frameworks to model the brain, paving the path for neural interpretability
Developing and proposing novel applications of AI to not only predict epileptic seizures, but enable us to achieve transparent insights into how epileptic seizures are caused and how they propagate through the brain
Ode 2 Socrates is a browser-based reflection tool
Built to help you look inward, question your assumptions, and form arguments with AI asking Socratic questions to challenge your thinking and interrogate your underlying assumptions.
It maps your thinking as a spatial graph, revealing connections and blind spots you didn't know were there.
"A space to find out what you think, how you think, and who you are."
Click here
Answering Biology GCSE questions can be hard. Exam-specific terminology means keywords and clear explanations are necessary, and generic AI tools often go too far in depth when answering biology questions
Fine-tuned SmolLM2-135M model using HuggingFace Transformers and PyTorch
Synthetically generated custom biology dataset adhering to A-Level and GCSE specification with labels
Implemented training checkpoints and mixed-precision training
Designed and implemented a novel neural network architecture incorporating Möbius strip topology to enhance feature learning and clustering
