Experience

AquaEasy (A Bosch Company)
• AI Data Scientist

  • Dec 2021 - Present


Singapore Institute of Manufacturing Technology

• Singapore
• Research
• 200 - 300 Employee

• Scientist

  • Mar 2018 - Dec 2021

  Jan 2021 - now Condition Based Predictive Maintenance for Production Line (Project Leader) • Analyse the log of machine downtime and maintenance, select useful information from the record of time series Alarm events. • Extract features from the alarm events using statistical methods, build Decision Tree and Isolation Forest models with the extracted features to detect abnormalities of the production line. Nov 2019 - Now Feature Analysis and Predictive Modelling for Semicon Manufacturing (Project Leader) • Extracted features from time series data using statistical method, ts-fresh and wavelet transform. • Built regression models with the extracted features to predict device remaining useful lifetime. • Applied Bayesian Optimization to fine tune the model configuration to improve the prediction accuracy. Feb 2021 - now Enhancement of Overall Equipment Efficiency Monitoring • Based on the sensor data collected from Injection Molding Machine, apply Random Forest and XGBoost model to estimate the machine state, normal or fail. Mar 2018 - Mar 2021 Industrial Internet of Things (IIoT) Hyper-parameter Optimization and Ensemble for Classifiers • Designed a new optimization algorithm based on the Conditional Neural Processes to automatically identify the best hyper-parameter setting for classifiers. • Evaluated the performance of six state-of-the-art ensemble methods which are used to unite a pool of classifiers that are tuned by hyper-parameter optimization. Oversampling for imbalanced classification • Designed a dual over-sampling strategy to improve the classification accuracy for imbalanced data. • Applied conditional Generative Adversarial Network(cGAN) and Synthetic Minority Over-sampling Technique(SMOTE) to increase the size of both majority and minority samples. Show less



Nanyang Technological University

• Singapore
• Higher Education
• 700 & Above Employee

• Postdoctoral Research Fellow

  • Apr 2016 - Jan 2018

  Research in resource management and optimization for distributed online applications. Oct 2017 - Jan 2018 Analyse the problem that many optimization algorithms are always trapped by the local optimum. If the local optimum has a broad basin around it and the global optimum is a small deep hole, most of the time and most of the algorithms will converge to the local optimum. Graph-based Online Game Data Analysis Aug 2017 - Oct 2017 1. Generated a large scale graph (more than 106 nodes and 109 links) to represent the social network of players in the online multi-player game, LOL. 2. Analysed some properties of the graph, such as degree distribution, clustering, assortativity, the largest connected component. Distributed Machine Learning Apr 2017 - Aug 2017 1. Surveyed the popular machine learning algorithms, like Topic Model (LDA), Lasso, Matrix Factorization and Distance Metric Learning. 2. Learned the two methods of implementing distributed machine learning: model parallel and data parallel. Theoretically studied the framework of parameter server, and understood its working process. Workload Forecast via Recurrent Neural Network Jan - Apr 2017 1. Investigated the efficacy of neural network based methods in predicting online application workload. 2. Proposed an intelligent strategy to estimate the prediction accuracies of various methods before knowing the real result. Cost Minimization in Amazon EC2 Apr-Dec 2016 1. Developed a fast and effective strategy to compute the optimal number of instances to acquire for three EC2 pricing options: on-demand, reserved and scheduled. 2. Designed a scheduling algorithm to arrange the scheduled instances in compliance with the restriction of their scheduled durations. Show less



Freie Universität Berlin

• Germany
• Research Services
• 700 & Above Employee

• PHD Student

  • Sep 2011 - Dec 2015

  Project: Restart in Mobile Offloading System 1. According to the system structure, designed stochastic models to simulate the system operation by using Petri-net based method. Synthetically combining task completion time, energy consumption and throughput as a metric to evaluate the efficiency of restart in network failure handling. 2. Utiliized stochastic model (Petri-nets) to simulate the system operation and evaluate its performance. 3. Developed a mobile offloading test-bed includes an Android App for the mobile client and a website project for the server. OCR is implemented as the sample offloading application. 4. Experimentally demonstrated the impact of unreliable network connection on the mobile offloading system. When the network quality is unstable, the unpredictable long delay in the offloading task completion is observed. 5. Theoretically designed several stochastic models to identify the optimal timeout for restart and experimentally evaluated its performance. 6. Proposed a restart algorithm by completing the offloading task locally in the mobile device to accelerate the task completion when the network connection is unreliable. The optimal timeout to launch local restart is identified by mathematical derivation from the probability density function of the task completion time based on a greedy method. 7. Designed a scheme to implement the restart algorithm into the real offloading test-bed. A dynamic histogram is proposed to follow the update of the distribution of the task completion time when the network quality is changed. Show less



State Key Laboratory of Information Photonics and Optical Communications

• Master Student

  • Sep 2008 - Apr 2011

  Project:OTN/WSON Simulation Software 1. Participated in the design and development of the software based on OTN/WSON protocols. 2. Simulated the routing, connection establishment and path protection protocols with the software. Project: Distributed Parallel Transmission in Core Optical Network 1. Proposed three kinds of resource distribution algorithm in GMPLS based WDM network. 2. Participated in building a distributed emulation test-bed (consist of 20 servers) for core optical network. Show less



IBM China

• 1 - 100 Employee

• Internship

  • Jul 2007 - Aug 2007

  Learned the basic hardware structure of server p595, understood its disaster back-up scheme. Assisted the expert engineers to install operation system for the servers(p series) and configured the system according to the client request. Learned the basic hardware structure of server p595, understood its disaster back-up scheme. Assisted the expert engineers to install operation system for the servers(p series) and configured the system according to the client request.