Kelvin Wey Han Chan
Data Scientist @Shell
Signup · Get unlimited contacts
WORK HISTORY
Data Scientist @Shell
London, GB
Designed and deployed end-to-end ML pipelines in AzureML for training, evaluating and quantising deep learning vision models, achieving 20× lower inference energy usage with minimal performance degradation.• Developed and deployed LLM-based agentic systems, implementing adaptive dialogue policies and tools to automate resolving 84% of routine customer interactions.• Implemented a robust evaluation framework for generative models, assessing response quality, safety constraints, robustness to distribution shift and resistance to prompt-based attacks. • Built GenAI pipelines for post-hoc model explainability, integrating unstructured text with structured time-series model outputs to produce interpretable summaries and market outlook analyses for stakeholders
EDUCATION
University of Cambridge
Master of Philosophy - MPhil, Advanced Computer Science (Natural Language Processing)
Methodist College Kuala Lumpur (MCKL)
Western Australia Certificate of Education
University of Malaya
Bachelor of Engineering - BE, Electrical
Catholic High School, Petaling Jaya
High School
ABOUT KELVIN WEY HAN CHAN
I am a data scientist passionate about applying Natural Language Processing techniques to understand the nuances of human communication. I am currently a Data Scientist at Shell, where I apply my NLP skills to a variety of projects to help accelerate the energy transition. About MeMy journey into the world of Natural Language Processing (NLP) began with a simple question: Why can\'t I talk to machines like I talk to people?This observation, sparked by the limitations of voice assistants, led me to delve into the field of NLP, where I discovered the intricate relationship between language, technology, and human emotions. This fascination propelled me into research, where I began to unravel the complexities of NLP and its potential to transform how we interact with machines.One aspect of language that particularly captured my interest was code-switching, the ability to seamlessly switch between languages mid-conversation. This phenomenon, prevalent in multilingual communities, allows us to express a wider range of emotions and connect with others on a deeper level. However, current language models struggle to understand and generate code-switched language, limiting their ability to capture the nuances of human communication.I believe that improving language models\' ability to handle code-switching is crucial for developing truly natural and engaging human-machine interactions. By enabling machines to understand and respond to code-switched language, we can create more empathetic and effective communication systems that reflect the richness and diversity of human language.I am committed to using my NLP expertise to make a positive impact on society. In my free time, I volunteer my skills to build NLP applications for non-profit organizations, helping them optimize their operations and reach a wider audience. I believe that technology, when used responsibly, can empower individuals and communities to achieve their goals and make the world a better place.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.