Greetings! I am a skilled machine learning researcher/engineer with over 5 years of experience in developing, implementing, deploying and maintaining machine learning models. I received my Ph.D. in Electrical Engineering from University of Saskatchewan in 2023. In the past 5 years, I was deeply involved in developing machine learning models for automated medical image analysis, disease diagnosis and image synthesis. I am experienced in processing a variety of medical image datasets, which include radiology, MRI and digital pathology images. I am a passionate individual who thrives to build and apply machine learning algorithms to solve real-world industry problems.

  • Location: Toronto, ON, Canada
  • Email: sherryqin1019@gmail.com

Interests

Foundation Models

Machine Learning

Computer Vision

Natural Language Processing

Software Engineering

Machine Learning Operations (MLOps)

AI/ML Algorithms

Image Processing

Experience

Toronto Metropolitan University

Sept 2023 - Present

Postdoctoral Researcher

  • Implement automated Whole Slide Images (WSIs) preprocessing pipelines using Python and Openslide.
  • Develop novel deep learning models for automated detecting and segmenting tumor cells in digital pathology images.
  • Build and maintain automated nuclei detection and segmentation models using various machine learning algorithms.
  • Optimize the model architecture and design to improve model robustness and efficiency.
  • Conduct comprehensive tests and statistical analysis to validate the model on in domain data and out-of-domain data.
  • Develop and deploy a web application that integrates various automated nuclei segmentation models using Flask and AWS SageMaker independently .

University of Saskatchewan

Sep 2018 - August 2023

Research Assistant/Ph.D. Student

Project 1: Automated thoracic disease localization system development

  • Worked on designing an AI model named disease decomposition network (DDN) to automatically locate, disentangle disease signs from the diseased chest x-rays (CXRs).
  • The developed model not only can detect disease signs but also synthesize a normal counterpart CXR for radiologists’ reference.
  • Conducted comprehensive tests and comparisons to validate the advances of the proposed model.

Project 2: Automated synthesis of medical imaging data

  • Investigated and evaluated the effectiveness of using generative adversarial networks (GANs) to augment the medical imaging data for better automated disease recognition.
  • Studied a more complex progressive growing GAN (PGGAN) for image synthesis. The experiments show that the generated images from PGGAN are in good quality and close to real medical images
  • Proposed and implemented a novel image synthesis network named semantically preserving adversarial unsupervised domain adaptation network (SPA-UDA) for overcoming the domain shift problem when applying pre-trained models on unseen CXR data, and the model performance improvements are significant.

Project 3: Automated speech - transcription alignment tool development

  • Collaborated with the professor at the department of linguistics in U of S to develop an automated speech - transcription alignment system (Aligner).
  • Used Kaldi ( the state-of-the-art automatic speech recognition (ASR) toolkit) and Shell scripts to customize the acoustic model successfully.

Project 4: Automated thoracic disease recognition/classification system development

  • Developed and implemented machine learning based methods for automated classifying thoracic disease in CXR.
  • One way is using handcrafted features with traditional machine learning methods (e.g., support vector machine (SVM), k-nearest neighbors (KNN), etc) to achieve this. The other way is using deep neural networks to extract features from images directly and classifying them in an automated manner. The classification accuracy is as high as 90%.
  • Investigated different approaches for addressing the imbalanced class problem in the dataset. Details can be found in the publication.

Yangtze Memory Technologies Co., Ltd

April 2018 - August 2018

Reliability Engineer

  • Worked on writing python scripts for extracting key data from logs of chip reliability tests.
  • Devised python scripts independently for automated running statistical analysis of the data from logs
  • Collaborated with colleagues to analyze the statistical data for improving the yield of flash memory chips.

Projects

  • All
  • Web-App
  • Research

Automated nuclei detection

Image recognition as Service

CXR Disease Classifier

CXR Disease Classifier

CXR Disease Localization and Disentanglement

CXR Image Generator

Skills

Languages and Databases

vectorlogo.zone vectorlogo.zone vectorlogo.zone upload.wikimedia.org vectorlogo.zone vectorlogo.zone

Frameworks

vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone upload.wikimedia.org

Tools

vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone vectorlogo.zone