Stephen Tai Email and Phone Number

• Plan and implement R&D projects with application of nanofiberso Create design specification and functional requirement specificationo Conduct functional tests and product verificationo Technology transfer to Operation: Create the SOP for Operation and train the personnel• Cooperate with CUHK in the study of biological and psychological effects in wearing masks during doing sports• Discuss with Operation about equipment requirements, production estimation and manufacturing plan• Coordinate with A&C, supplier and OEM for making WIs of QC methods which comply to relevant industrial standards• Application of ITC research fundings open for industry• Coordinate procurement of Nanospider™ NS 8S1600U and determine the values in its acceptance test• Assist in formation of R&D team, admission into HKSTP and planning for the new laboratory• Integrate R&D section into current ISO13485 Quality Management System

• Project: Classification & Feature-based image compression of multi-ATOM cell images using CNNs• Programming languages: Python (include packages of numpy, scikit-learn, panda etc) & PyTorch • Classification on 7 types of lung cancer cells and 2 types of breast cancer cells• Analysis of latent space in U-Net (for understanding the features learnt by network)• Customize the parameters of U-Net for optimizing the performance• Automatic removal of image background for better feature extraction of the cells• Side project: Auto-focusing of microfluidic chip• Use SQL as a log book for CNN training

• Project: Development of automatic 3D segmentation algorithm of brain lesions (black holes) on T1-weighted (T1w) MR images of multiple sclerosis patients• Programming language: MATLAB, Python (include packages of numpy and panda) & PyTorch• Dynamic threshold approach: Identify T1 lesions by a threshold based on FLAIR lesions and local variable intensity• Deep learning approach (self learning via online lectures): Training of model with U-Net• Side project: Segmentation of total intracranial volume on T1w images using U-Net → Result: Better than conventional method

I worked on the project about building a Graphical User Interface (GUI) for Sub-bottom Profiler. Sub-bottom Profilers identify the sediment layers below the seabed by analysing the reflected echoes. GUI provides a platform for displaying the signal as well as allowing user to process the signal for the extraction of important features. MATLAB GUIDE was used in this project.

Master Thesis: Algorithm development for simulation and experimental validation of Ultrafast Doppler Imaging. This project targeted at analysing flow velocity in blood vessel. MATLAB and Field II Simulation Program were the programming languages used in this project. The largest advantage of Ultrafast Doppler Imaging, based on compound plane wave imaging, is that it is able to obtain quantitative velocity information in each pixel of Doppler image. It is because an ultrafast frame rate of up to several thousands Hz could be obtained when plane wave is used for imaging.

Our company (NAMI), which is funded by Hong Kong Government, cooperates with industry for developing the existing scientific technologies into commercial products. My major task was product development and optimisation, but also responsible for technology transfer, industrialising the production, explaining the results to clients and brainstorming new ideas for other potential products. Our team focused on needleless electrospinning using Elmarco Nanospider. The commercial product already available on the market is NASK Smart Mask.

The research group focuses on photoacoustic imaging and tomography. In the three months of my internship, I was responsible for two projects, i.e. absolute temperature measurement and respiratory gating. Thanksful to the research group, two journal articles were published at later days.