Current work
NUS I-FIM · Research Fellow
Delta learning for
materials discovery.
I work on delta machine learning within the NRF Materials Data Foundry (MDF) project at the Institute for Functional Intelligent Materials, National University of Singapore.
The MDF is a collaboration between NUS I-FIM and the University of Toronto's Acceleration Consortium. It aims to connect synthesis routes, measurements, and simulations in a unified materials dataset, supporting AI-guided discovery.
My contribution focuses on delta machine learning. This connects my background in quantum chemistry and computational workflows with the project's broader materials discovery programme.
What is delta machine learning?
Instead of learning a target property from scratch, a model learns the correction between a lower-cost baseline and a higher-fidelity reference. A prediction combines the baseline with that learned correction.
Higher-fidelity estimate = baseline prediction + learned correction
Its usefulness depends on the reference data, the baseline method, and whether the model generalizes to the materials being studied.
Foundational Δ-ML paper ↗Excited-state photophysics
Understand the mechanism.
Resolve how molecular structure and conformation govern emission, electron transfer, and non-radiative decay.
First-author work on restricted TICT and aggregation-induced emission.
DFT / TD-DFT · CASSCF / CASPT2 · Nonadiabatic dynamics
Related publication →Fluorophores & photostability
Design for function.
Translate photophysical mechanisms into design principles for imaging probes, photoswitches, and photoactive materials.
Collaborative studies in Nature Methods, Nature Communications, and Angewandte Chemie.
PET · TICT · ESIPT · Structure–property relationships
Related publication →What I bring to a collaboration
Calculations that help
answer the chemistry.
My strengths span electronic-structure calculations, interpretation of excited-state pathways, and the translation of computational results into molecular design questions.
- Investigating photoinduced electron transfer, twisted intramolecular charge transfer, and excited-state proton transfer.
- Relating structure and conformation to emission, photostability, and non-radiative decay.
- Building reproducible Python workflows for quantum chemical calculations, output analysis, and property prediction.
- Working with experimental colleagues to connect computed mechanisms with spectroscopy and imaging.
Research resources
A reproducible workflow for PET.
Computational data and source code for a DFT/TD-DFT workflow investigating photoinduced electron transfer in fluorescent molecules, including inputs, outputs, and analysis scripts.
Explore the Zenodo record ↗Software
Quantum Chemistry Input Generator
A tool for preparing quantum chemical input files and supporting repeatable computational workflows.
View the repository ↗