Ai Engineering Apprentice
Current• Collaborated with a team of researchers and doctors at a leading local hospital to enhance a transformer-based drug response prediction model for cancer treatment. • Utilised advanced biomedical knowledge graph embedding approaches to generate specialised, context-aware embeddings that deepened insights into drug-gene interactions. • Conducted comprehensive evaluations of various SOTA heterogeneous graph embedding techniques. Reduced graph noise and derived context-specific subgraphs through n-hop neighbourhood and meta-path filtering. Successfully improved embedding quality by 80% (Hits@10), while also achieving substantial efficiency gains.• Refactored a novel graph representation learning methodology inspired from subword tokenisation strategies commonly used in the NLP domain to overcome an intricate gene token mapping process.• Developed a RAG-based chatbot powered by GPT-3.5 Turbo to assist migrant workers with limited English proficiency. Leveraged APIs and libraries like LlamaParse, LangChain, Azure OpenAI, and Chroma (vector database). • Led and facilitated mentoring sessions on topics such as Deep Neural Networks and Computer Vision, as well as the core principles behind generative models like Stable Diffusion.
