Starting from a random phase mask, perform FFT and IFFT back and forth. During this procedure, enforce the measured amplitudes from the image plane and the Fourier plane. The phase retrieval happens because the results of FFT and IFFT are in complex form, so by considering the real and imaginary parts of the result, we can update the phase.
Developed an adjoint-based inverse design framework to optimize nanoscale optical masks for diffraction control, leveraging Abbe’s imaging theory and Fourier-domain constraints.
OLED luminance degradation prediction over future time windows using a transformer-based time-series model. Delta luminance was incorporated to improve training stability and forecasting performance.
Exploring acousto-optic effects in air to generate controllable refractive index perturbations for free-space display systems.
Reducing hallucinations using lightweight retrieval-augmented generation (RAG).
Can we inversely predict mask geometries to achieve accurate patterning beyond diffraction limits based on Hopkins/Abbe imaging theory?
Image generated by ChatGPT. 
Precise nanoscale patterning enables strong electric-field enhancement at individual resonators. 