1. Proof of concept of the operator autoscaling in SAP Data Intelligence pipeline modeler.2. Develop and deploy a Helm chart of an open-source Kubernetes metric adapter in the cloud data management platform SAP Data Intelligence.3. Use Kubernetes Horizontal Pod Autoscaler to scale up/down the computing resources based on several custom metrics.

1. Proposed a novel Generative Adversarial Network based on self-attention to enhance the satellite earth observation data.2. Implement the above deep learning model with Tensorflow and scale it up to high-performance computing (HPC) system with Horovod. State-of-the-art performance is achieved on several standard metrics.

1. Developed an artificial intelligence web-demonstrator, which can show the entire procedure to apply machine learning to predict the cost of some machine tools (including preparing training dataset, pre-process data, model training, and prediction).2. Develop a mix-reality application on Microsoft Hololens to scan QR codes on machine tools, then request and display information about those tools (e.g. production progresses, technical configurations, etc) from a database. A collaborative project with WBA Aachener Werkzeugbau Akademie GmbH.

1. Develop a login page for a vocabulary publishing website2. Add user authentication functionality to the backend database3. Proof of concept of a vocabulary recommendation algorithm

Participate in a cross-media retrieval project, design and implement a deep model to extract a common representation of data with different media types (e.g. text, image, video, voice, and 3D shape), which is suitable to the scenario where query and retrieval results are of different media types.