I am Udith Haputhanthri, a Post-Baccalaureate Fellow affiliated with Wadduwage Lab, Center for Advanced Imaging at Harvard University. I am fortunate to be advised by Dr. Dushan Wadduwage and co-advised by Dr. Sergey Ovchinnikov. I work at the intersection of microscopy and deep learning/ computer vision and biology-inspired artificial intelligence (AI).
In my recent work, my colleagues and I have been developing a $\sim$ x100 faster quantitative phase microscopy using learnable optics and computer vision. The proposed optical setup selectively extracts important features from the phase distribution of the light field that comes from a biological specimen (through diffractive deep neural networks). A conventional photodetector array then captures the extracted feature representations. Finally, super-resolution algorithms reconstruct the phase distribution of the input field.
In parallel, I have been exploring neuroscience/ cognitive science/ biology-inspired AI. My goal is to understand how biological brains learn and use those insights to improve AI. I am currently working on biology-inspired optimization algorithms to optimize non-overparameterized neural networks.
I received my B.Sc. degree in Biomedical Engineering from the University of Moratuwa, Sri Lanka 2022 (1st class, cGPA: 4.00/ 4.20). Prior to joining as a post-baccalaureate fellow at Harvard, I worked at the same lab as a Visiting Undergraduate Research Fellow (Remote) in 2021/ 2022. I designed a fully differentiable computational optical model for fluorescence microscopy. The proposed computational model allows high-throughput imaging by learning content and task-aware illumination patterns.
I have also been working as a machine learning/computer vision researcher (part-time) in the Biomedical Research and Innovative Collective (theBRIC) at the University of Moratuwa, Sri Lanka, and The AI Team. I completed my internship in Acceler Logic, where I contributed to develop 5-bit neural network quantization algorithms for high-throughput analog hardware. The project was outsourced by Analog-Inference, California, USA.
Outside of work, I was a chess player who loved rook endings and I am an acoustic guitarist who skipped music classes. In addition, I have been learning piano at a snail’s pace and I love free-style wrestling.
|Jul 2022||Started work as a Post-Baccalaureate Fellow in Center for Advanced Imaging at Harvard|
|Jun 2022||Deep optical coding design in computational imaging: preprint available on arxiv|
|May 2022||From Hours to Seconds: Towards 100x Faster Quantitative Phase Imaging via Differentiable Microscopy: preprint available on arxiv|
|Apr 2022||Serving as a reviewer for ECCV 2022|
|Mar 2022||Differentiable Microscopy Designs an All-Optical Quantitative Phase Microscope: preprint available on arxiv|
|Mar 2022||Differentiable Microscopy for Content and Task Aware Compressive Imaging: preprint available on arxiv|
|Mar 2022||Complete paper is submitted to IEEE Signal Processing Magazine, 2022|
|Feb 2022||White paper “Learnable Diffractive Optics for Efficient Imaging Systems” got accepted to IEEE Signal Processing Magazine, 2022|
|Jan 2022||Did a talk ”Machine Learning in Action”, Informatics Institute of Technology, Sri Lanka|
|Nov 2021||Serving as a reviewer for CVPR 2022|
Deep Optical Coding Design in Computational ImagingArXiv 2022
From Hours to Seconds: Towards 100x Faster Quantitative Phase Imaging via Differentiable MicroscopyArXiv 2022
[P]Differentiable Microscopy Designs an All Optical Quantitative Phase MicroscopeArXiv 2022
[P]Differentiable Microscopy for Content and Task Aware Compressive Fluorescence ImagingArXiv 2022
[C] ICASSPTowards Accurate Cross-Domain In-Bed Human Pose EstimationArXiv 2021
[C] SPIE BiOSDEEP learning powered De-scattering with Excitation Patterning (DEEP) for high-throughput wide-field multiphoton microscopyIn Multiphoton Microscopy in the Biomedical Sciences XXII 2022
[C] ISCASA Novel Transfer Learning-Based Approach for Screening Pre-Existing Heart Diseases Using Synchronized ECG Signals and Heart SoundsIn 2021 IEEE International Symposium on Circuits and Systems (ISCAS) 2021
[J] IJNMETopologically optimal design and failure prediction using conditional generative adversarial networksInternational Journal for Numerical Methods in Engineering Dec 2021
[J] JACMNonlinear Multiscale Modelling and Design using Gaussian ProcessesJournal of Applied and Computational Mechanics Dec 2021