Research
My research interests sit where large pretrained models meet robot control. Below is what I want to work on, what I've done that leads there, and the coursework behind it.
01
Policies that take in camera images and a written instruction and output motor commands directly. I want to understand how well these models ground language in what a robot actually sees, and how far they transfer to a new robot with little data.
02
Learning control from trial and error. The part I find most interesting is getting a policy trained in simulation to work on a physical robot, and whether pretrained models can cut down how much interaction that takes.
03
Most of my hands-on work so far. Perception, state estimation and control running on real hardware, plus the electronics underneath. This is the foundation I want learned policies to sit on.
Research experience
Dec 2024 to May 2025
Singapore
National University of Singapore · MARMoT Laboratory
PCB design · Robot control · Embedded systems · Research
Dec 2023 to Sep 2024
Sri Lanka
Qubebots · Xfly drone platform
Computer vision · Drones · Localization · Embedded
Reproductions & research projects
ECCV 2024 · Reproduction
Mixture density networks that forecast multimodal pedestrian trajectories with calibrated uncertainty, for autonomous applications that must plan around people.
Autonomous racing
Landmark-based Extended Kalman Filter SLAM for a differential-drive robot navigating cone-marked race tracks, mapping and localizing simultaneously.
CVPR 2025 · Implementation
Paper implementation of momentum-aware planning for end-to-end autonomous driving, keeping planned trajectories temporally consistent.
Coursework
Contact
If your lab or team works on VLA models, reinforcement learning or physical AI, I would be grateful to hear about any research positions or collaborations. Please feel free to write to me.