Exploring AI safety & Physical AI
- Currently working on research in AI safety evaluation, VLA/RL & robotics manipulation, model optimisation/distillation and related areas while exploring opportunities to build or join a startup.
Grokking & Hillclimbing · Based in Malaysia
I'm competitive by nature, agentic by habit.
Coming from a non-linear and under-resourced background, I started my journey at age 10 by teaching myself to hack a game website and jailbreak phones and eventually found my way to frontier AI research. And that itself has trained me to be highly adaptable at solving and learning from unfamiliar and hard problems, and most of the time, I'm obsessed with problems I truly care about.
I'm currently building on AI safety evals for LLMs and physical AI, mechanistic interpretability, and model inference optimisation/distillation. My research interest lies in understanding how the models work internally at the layer level, and how that understanding can push AI forward while keeping it safe and fast enough to deploy in real-world settings.
Places I've worked in: EleutherAI,
UniverseTBD, Shanghai Jiao Tong Uni.,
Uni. of Maryland, Metacreation Lab,
SEACrowd.
Selected work