Research Program
Developing Robust, Trustworthy, and Efficient AI for Perception and Real-world Deployment
Embedded Files
My research program focuses on developing robust, trustworthy, and efficient AI, grounded in signal processing, for perception and real-world deployment. While modern AI has achieved remarkable success, real-world deployment requires addressing key challenges, including robustness under distorted inputs and distribution shift, learning from limited and biased data, and ensuring privacy against risks such as model inversion.
To address these challenges, my work is organized around a unified research program that advances principled algorithm and system design under data, model, and deployment constraints.
Figure: Overview of my research program, integrating three methodological pillars to enable robust, trustworthy, and efficient AI models and systems for real-world deployment.
Selected Highlights
Selected Highlights
200+ publications in leading venues including IEEE TIP, NeurIPS, CVPR
(Google Scholar for updated and full publication list)
Top 1% most cited paper in IEEE Transactions on Image Processing
CVPR Best Paper Finalist
Finalist for the Super AI Leader (SAIL) Award at the World Artificial Intelligence Conference (WAIC)
Lead PI of multi-million-dollar competitive research grants
14 U.S. patents supporting real-world deployment of image processing and AI systems
Research establishing the technical foundation for KroniKare, an AI healthcare solution deployed in hospitals
Nearly a decade of collaboration with DSO National Laboratories with sustained technology transfer to defence applications
Clinical collaborations at the national healthcare system level with MOHT, SingHealth, and NNI
Robust and Trustworthy AI
Ensuring reliability, robustness, and privacy in AI under real-world conditions.
Robust perception under noisy and uncertain conditions
Adversarial robustness and generalization
Privacy and model inversion in deep learning
Data-Efficient Learning and Generative AI
Enabling learning under limited, biased, and shifting data distributions.
Learning under limited and biased data
Distributional robustness and fairness
Generative modeling under data constraints
Compute-Efficient AI
Designing AI that operate efficiently under real-world resource constraints.
Efficient model design and optimization
Deployment under real-world resource constraints
Scalable AI for practical applications
Real-World Deployment and Impact
My research has led to real-world deployment across multiple domains, including healthcare, defence, sustainability, and industry, translating advances from my research program on robust, data-efficient, and compute-efficient AI into practical systems operating under real-world constraints.
My work established the technical foundation for KroniKare, a SUTD spin-off founded by my former postdoctoral researcher, where I provided mentorship and early-stage guidance on technology translation and industry engagement. The resulting technologies have been deployed in hospitals and piloted in Europe.
This line of work has further expanded into clinical collaborations at the national healthcare system level, including partnerships with MOHT, SingHealth, and the National Neuroscience Institute (NNI), supporting the development of AI systems for dementia and neurodegenerative disease using real-world longitudinal data.
In parallel, my research has led to sustained collaboration with DSO National Laboratories over nearly a decade, resulting in the transfer of algorithms, models, and technical know-how into national defence capabilities.
I also lead ongoing industry collaborations under my project HyperSense.ai, focusing on AI for waste auditing and sustainability through material-aware and data-efficient AI systems.
Page updated
Google Sites
Report abuse