Anni Huang

During the internship, I finished two projects. The first is predictive maintenance for diamond cutter machines. It’s a research project with Macaw Technology University. I am responsible to conduct experiments and interpret the results. And its main problem is a binary classification problem. But I also tried to combine unsupervised learning, regression in this project. Since their result is not good enough to use in production, our final model is limited to the binary classification of whether there will be alarm after t_pred, and t_pred is changeable. In production, we want the as large as possible. So, we set it to be 2 hours since the model can have more than a 95% recall rate (this is a hard requirement for models to be used in production) and engineers can have enough time to fix the issues. Another is predicting the temperature given the temperature and warpage of the current production batch, which is called MOCVD in short. This project is like the optimization of a previous project. In the past, our company used empirical formulas to adjust temperatures. But now, it wants to try to use machine learning algorithms to make better adjustments. I solved it by using KNN to find nearby points and use regression models to improve on it. During this internship, I exploded a lot about different tools and techniques in data mining. My job scope is to support the product and innovation team to conduct experiments to test their ideas and report to Shenzhen and Singaporean teams about my findings.