This project introduces EchoPal, a prototype system that allows a human receptionist to converse with the user synchronously via an Amazon’s Echo device without access to the user’s voice recordings. EchoPal (i) make sense of noisy automatic transcriptions (ii) quickly generate suggested responses to provide timely help and (iii) support quick search functions to handle complicated queries. We study end-to-end system latency and conversation quality with several pairs of participants, overall quality score is 3.71 / 5.0 from 17 participants and 88% of users considered it outperformed than or similarly good as Amazon’s Echo devices. Overall, our work demonstrates the challenges and potentials of human-in-the-loop voice-enabled devices, which opens up new possibilities for future smart devices and smart environments.

This project built a web-based automatic crowdsourcing system to allow users to upload food photographs and receive correct restaurant information (90% accuracy) within a few minutes. In addition, the restaurants information is generated from a series of tasks on Amazon Mechanical Turk.
[PPT] [GitHub]

This project created a website which allows the user to check the school policy of COVID-19. The information includes school name, location (State), when school started virtual learning; when the virtual learning will be stopped, any support for international students, etc.

This project aims to predict total sales for every product and store of a large Russian software firm using Machine Learning techniques. We worked with a challenging time-series dataset consisting of daily sales data and implemented several advanced algorithms as XGBoost, Long Short-Term Memory (LSTM), Random Forest and Convolutional neural network (CNN) on this dataset. Currently, our final prediction result is ranked top 15% on Kaggle leaderboard.
[GitHub]

This project implemented the statistical testing (McNemar’s test and 5x2cv paired t test) by Python and analyzed the experiments on four different machine learning classifiers with the wine dataset.
[PDF] [GitHub]

• Managed several projects (~$20 million) in advance metrology field and smart manufacturing area.
• Planned and analyzed cost control and project scheduling; Performed risk assessments, strategic planning, and communications across customers and team members.

• Interpreted global trading data, policies, economic, and technology trends of the machine tool industry to make recommendations and offer counsel to the Taiwan government.
• Organized technical exhibitions and forums in Thailand, Indonesia, and Russian to bridge domestic machine tool manufacturers and foreign potential customers.
[Articles Link_1 (Chinese Version)] [Articles Link_2 (Chinese Version)]

• Productized photovoltaic technology for smart energy management: developed new photovoltaic ceramic tiles to boost efficiency of electricity generation with a reduction of building's energy consumption up to 26.8 % and transferred to industry successfully for mass production.
• Established a Building Integrated Photovoltaic database: modeled and simulated complex solar systems and environmental factors with large datasets for a 90% accuracy forecast of power generation and consumption.
[Paper_1] [Paper_2] [Paper_3 (Chinese Version)]