Sam He

I was graduated from Master of Data Science in University of Sydney with WAM 72.6 and 80.3 for related graduate certificate. Achieved High distinction on Data visualization and Distinction on natural language processing as well as the capstone research project. Contributed as team leader on the capstone project “Identification of Cancer Risk Groups by Integrative Multi-omics Using Autoencoders and Tensors”, successfully detected high/low risk patients for cancer disease using multi-omics data. Besides the knowledge gained from campus, I am passionate to develop my own project, join IT industry meetup or solve Leetcode problems to keep myself refreshing on IT/Data area.

Degree related projects

Seminal management windows application

C#, .Net, MySQL, Visual Studio 2017

• Contributed as programmer in the team, building up function related to seminal management including create/edit/remove/highlight seminal as admin and user.
• Assist on designing and building part of the Sql database which is related to seminal data.

Full stack website for Training booking and management.

Angular, NodeJS, ExpressJS, MongoDB, AWS, Visual Studio Code

• Design a training schedule website for a gym with functions securely sign up/in/out, view/book/manage training timetable, create/manage/edit users with different accessibility. (Talking about my own achievement with the result instead of introducing the project)
• Build a training management application with functions including securely sign up/in/out, view/book/modify training timetable.
• Be responsible on building a real-time chatting function and simple email server allowing user communication by Socket.io and Nodemailer.
• Web application deployment via heroku and AWS.

Name Entity Recognition from the defence corpus

Python, Pytorch, Google Colab 

• Design and concatenate different input embeddings including syntactic, semantic textural features(pos tag, glove, etc) to fulfill the elements of input data.
• Combining multi-layers Bi-LSTM and CRF as the base NER model to attach the name tag for each data. Also design my own attention method to boost the performance of the model, which result in 84.3 Mean F1 Score, 4% increased by the baseline model.
• Using of more than 30 pairs of experiment data to design ablation study and evaluation plan. Generate an professional academic report with the experiment as well as visualizing experiment data. Distinction rewarded.

Identification of Cancer Risk Groups by Integrative Multi-Omics using Autoencoders and Tensors

Python, R, Jupyter Notebook, Google Colab, R studio

• Work on the multi-omics data with limited samples under Agile software development methodology and complete the academic paper of the project submitting to ACM Health.
• Design own methodology using tensor decomposition and autoencoder to solve the bottleneck about compressing the huge original genomics data.
• Supply survival analysis/tumor purity classification on the compressed and decomposed omics data to identify risk level for each group of patients, achieved using R and generate the statical report.