I'm a full stack Machine Learning Engineer with an extensive research background and experience spanning the entire ML pipeline - from data analysis and preprocessing to model development, optimization and deployment.

My current areas of interest include LLM customization and cloud computing for generative AI.

In my job, I am developing a tool that enhances architects' creative workflows by integrating generative AI and computational tools. In my free time, I am exploring an LLM-based approach to building a competitive AI for Pokémon Showdown, and have developed a working early prototype.

I have been building applications that use LLMs for the past year. AGI pilled. when i was in uni i participated in about 11 hackathons and won a 2 which is not a good rate but i learnt a lot.

Agents

automatic financial statements aggregator that takes all your bank and credit card statements of that month and presents clear insight to where your money is going A recommendation algorithm that you control .

Clarence Lee Sheng is a Research Engineer at DSO National Laboratories, specializing in LLM Pretraining. He holds degrees in Design and Artificial Intelligence from Singapore University of Technology and Design (SUTD) and Raffles Institution, bringing 4 years of experience to his role.

Pretraining large-scale multilingual LLMs with billions of parameters and trillions of data points at DSO National Laboratories. Additional work includes using Unity3D to bridge vision and language, and researching uncertainty estimation and calibration techniques for language models.

I’m an AI research engineer at DeCLaRE Lab, where I work on improving RAG and reasoning systems. My recent work includes evaluating LLM trustworthiness and improving model grounding (accepted ICLR 25). In a past life, I researched on cancer biology and brain-computer interfaces, with published work in the field. Beyond research, I enjoy building projects that pique my curiosity or automate everyday tasks—like search systems and slot-booking bots. Fun fact: I won a scholarship to Stanford’s summer school, expecting a typical study trip. Instead, I watched a meteor shower on a mountain, hitchhiked from a remote beach, and even tried launching a startup. It sparked my love for reading, meeting new people, and embracing uncertainty.

I am interested to learn more about scaling search in production, the challenges they face and how they solve them. I am also interested in connecting with those passionate for chain of thought reasoning and agentic workflows.

I am currently working on building a mini search engine (https://github.com/shanghongsim/retrieval_prototype). I am currently testing various retrieval methods and foudn that reranking works best! I am planning to try implementing K means clustering before reranking to improve efficiency. In the next stage, I am planning to make (1) a web scraper to scrap relevant websites, (2) try hosting a chat LLM in AWS to be able to give AI responses and (3) wrap with a frontend to showcase the results!