Singapore
Machine learning (ML) researcher with over 4 years of experience. Proven skill in ML research and end-to-end ML. Track record of (1) building end-to-end ML agentic retrieval augmented generation (RAG) system and (2) deploying multimodal ML system on cloud platform. Published ML research in top-tier peer-reviewed journals. Gained international research and development experience at Israeli start-up. Seeking to leverage ML expertise to solve real-world problems.
• Built an end-to-end agentic RAG system in the form of an internal AI assistant support prototype platform for enterprise use. Developed agentic AI with LangChain. Implemented vector database with FAISS. Monitored performance using MLflow. Deployed on Google Cloud Run. GitHub, and API links are available in the “Projects” section under "End-to-end Agentic RAG System for Enterprise Support". • Fine-tuned and deployed a vision-language transformer for ID card fraud detection. Optimized hyperparameters with Optuna. Containerized with Docker. Built backend using FastAPI. Deployed on Google Cloud Platform (GCP). Built an interactive Gradio demo on Hugging Face Spaces. GitHub, API, and demo links are available in the “Projects” section under "Deployment of Vision-Language Transformer for Fraud Detection". • Led a novel deep learning study to identify scenarios where pre-trained models succeed and fail in improving downstream Alzheimer’s disease prediction performance. Improved model performance by 20%. Developed models using PyTorch. Findings were published in the Q1 journal Human Brain Mapping. Paper is available in the "Projects" section under "Evaluation of Foundation Models in the Transfer Learning Context". • Spearheaded a novel simulation study to optimize image pre-processing for brain stimulation treatment. Simulations performed via Python, Pandas and Numpy. Improved precision of stimulation target by 30%. Findings are published in a pre-print server while manuscript is undergoing peer review. Paper is available in the "Projects" section under "Computational Simulations to Optimize Processing of Images".
• Designed an automated large-scale data pre-processing pipeline to process over 100,000 brain images. Programmed pipeline using Bash and MATLAB on a Linux system. Implemented pipeline on a high-performance computing (HPC) cluster. • Developed interpretable machine learning models to understand key predictors of depression. Collaborated with a non-technical medical doctor. Implemented models using Scikit-learn. Paper is available in the "Projects" section under "Explainable AI for Predictive Feature Analysis". • Led a novel ML study to use time-series data from wearables to predict future cognition. Performed exploratory data analysis (EDA) using Pandas and SciPy. Developed a XGBoost model.
• Two Birds Medical Pte. Ltd. was a medical device start-up. • Cooperated with a medical doctor to perform market research for surgical device.
• MobileODT was an Israeli start-up. • Created a microcontroller and electrical circuit board prototype for the lighting system of the next-generation imaging system for Israeli startup, within 3 months.
• NUS MedTech was an undergraduate student society. • Led a diverse cross-faculty team of 10 students to build a new undergraduate student society, for launching medical technology startups.