I am currently studying for a Master's degree in the Institute of Computation and Modeling Science (ICMS) at National Tsing Hua University. My research interests are in AI, object detection, defect detection, and Data science. Prior to this, I worked as a data scientist in the data department of AVATACK CO., LTD., where I was incharge of the Data Lake, data analysis, AI research and applications.
https://github.com/leoliu5550 https://www.linkedin.com/in/leo-liu-aa83721b3/ [email protected]
Institute of Computation and Modeling Science
Department of statistics and information Science
清華 • Sept 2022 - Present
I played a key role in constructing and optimizing an embedded object detection model with an 86% recall rate on the i.MX RT1060EVK, enabling real-time identification of people and taxis in a streaming monitor, enhancing traffic efficiency and minimizing vehicle idling.
Jubo Health • Jul 2023 - Sept 2023
The internship focused on creating a Conversational User Interface (CUI) for a voice input system to streamline nurses' indirect care tasks, particularly administrative paperwork, aiming to enhance note-taking efficiency and minimize human errors. It estimated a reduction of nurses' note-taking time by 1 hour in one shift.
Symbio, Inc. • Jun 2022 - Jul 2022
As an OAC consultant at BenQ, I specialized in guiding users to master Oracle Analytics Cloud for effective data visualization. I also set up vital data infrastructure to optimize user operations and created and delivered 15 personalized tutorial courses tailored to user needs.
Symbio, Inc. • Oct 2019 - May 2022
As a data scientist at Four-Dimensional Innovative Materials, I specialized in governance and data integration analysis for the group's data platform, with responsibilities spanning optimization of production process conditions, predictive equipment maintenance, and upkeep and enhancement of the Business Intelligence (BI) system.
Mathematical Modeling and Analysis-Lotka-Volterra equation analysis
Topics on Applied Mathematics - YoloV3 on IPCam DataSet(on tensorflow2)
Special Topics on Scientific Computing - YoloV7/DDN/DERT on AOI Dataset(on pytorch)
The Study of User and Chatbot Design - wound detection&chatgpt linebot on aws service
2023 Mei-chu Hacksthon
In the i.MX RT1060EVK, object detection is embedded to detect the number of people and taxis, which increases traffic convenience and reduces the idling time of vehicles. I was responsible for building and quantizing the object detection model, which can detect all people and taxis in a streaming monitor.
The real-time model's recall achieved 86%. SLIDES, picture reference
Jul 2023 - Sept023
Taipei, Taiwan
The internship aimed to develop a Conversational User Interface (CUI) for a voice input system, intending to alleviate the time spent by nurses on indirect care tasks, such as administrative paperwork. The goal was to eliminate the limitations nurses face in note-taking during care, thereby reducing the occurrence of human errors in data generation.
It estimated a reduction of nurses' note-taking time by 1 hour in one shift.
The internship involved collaboration with the National Taiwan University (CAE) and received recognition in an internship competition.
Jun 2022 - Jul 2022
Taipei, Taiwan
Oct 2019 - May 2022
Taipei, Taiwan
After completing my Statistics and Information Science degree, I began my career as a data scientist in the core data department at FOUR-PILLARS. I was responsible for optimizing production process conditions, implementing predictive maintenance for equipment, and maintaining and developing the company's Business Intelligence (BI) system.
Project as Data Scientist
in 2019~2022
As certain products necessitate manual inspection, this process is susceptible to individual subjectivity, variations in training, fatigue, and distractions. These factors contribute to inconsistent inspection standards.
MAIN RESULTS:
Due to the inherent nature of the product properties, all product testing is conducted through destructive methods, making it impractical to test each individual product. However, by gathering sensor data within the production facility, we can leverage machine learning technology to train an AI model.
MAIN RESULTS:
Designing a data model and creating dashboards for Sales, Finance, and Production on Oracle Analytics Cloud has helped project-related personnel save 3-4 days per month.
Create a data pipeline that integrates Manufacturing Execution System (MES), Enterprise Resource Planning (ERP), and sensor data within the production facility. This pipeline will assist production technicians in monitoring products along the production line and preparing for AI applications.
I am currently studying for a Master's degree in the Institute of Computation and Modeling Science (ICMS) at National Tsing Hua University. My research interests are in AI, object detection, defect detection, and Data science. Prior to this, I worked as a data scientist in the data department of AVATACK CO., LTD., where I was incharge of the Data Lake, data analysis, AI research and applications.
https://github.com/leoliu5550 https://www.linkedin.com/in/leo-liu-aa83721b3/ [email protected]
Institute of Computation and Modeling Science
Department of statistics and information Science
清華 • Sept 2022 - Present
I played a key role in constructing and optimizing an embedded object detection model with an 86% recall rate on the i.MX RT1060EVK, enabling real-time identification of people and taxis in a streaming monitor, enhancing traffic efficiency and minimizing vehicle idling.
Jubo Health • Jul 2023 - Sept 2023
The internship focused on creating a Conversational User Interface (CUI) for a voice input system to streamline nurses' indirect care tasks, particularly administrative paperwork, aiming to enhance note-taking efficiency and minimize human errors. It estimated a reduction of nurses' note-taking time by 1 hour in one shift.
Symbio, Inc. • Jun 2022 - Jul 2022
As an OAC consultant at BenQ, I specialized in guiding users to master Oracle Analytics Cloud for effective data visualization. I also set up vital data infrastructure to optimize user operations and created and delivered 15 personalized tutorial courses tailored to user needs.
Symbio, Inc. • Oct 2019 - May 2022
As a data scientist at Four-Dimensional Innovative Materials, I specialized in governance and data integration analysis for the group's data platform, with responsibilities spanning optimization of production process conditions, predictive equipment maintenance, and upkeep and enhancement of the Business Intelligence (BI) system.
Mathematical Modeling and Analysis-Lotka-Volterra equation analysis
Topics on Applied Mathematics - YoloV3 on IPCam DataSet(on tensorflow2)
Special Topics on Scientific Computing - YoloV7/DDN/DERT on AOI Dataset(on pytorch)
The Study of User and Chatbot Design - wound detection&chatgpt linebot on aws service
2023 Mei-chu Hacksthon
In the i.MX RT1060EVK, object detection is embedded to detect the number of people and taxis, which increases traffic convenience and reduces the idling time of vehicles. I was responsible for building and quantizing the object detection model, which can detect all people and taxis in a streaming monitor.
The real-time model's recall achieved 86%. SLIDES, picture reference
Jul 2023 - Sept023
Taipei, Taiwan
The internship aimed to develop a Conversational User Interface (CUI) for a voice input system, intending to alleviate the time spent by nurses on indirect care tasks, such as administrative paperwork. The goal was to eliminate the limitations nurses face in note-taking during care, thereby reducing the occurrence of human errors in data generation.
It estimated a reduction of nurses' note-taking time by 1 hour in one shift.
The internship involved collaboration with the National Taiwan University (CAE) and received recognition in an internship competition.
Jun 2022 - Jul 2022
Taipei, Taiwan
Oct 2019 - May 2022
Taipei, Taiwan
After completing my Statistics and Information Science degree, I began my career as a data scientist in the core data department at FOUR-PILLARS. I was responsible for optimizing production process conditions, implementing predictive maintenance for equipment, and maintaining and developing the company's Business Intelligence (BI) system.
Project as Data Scientist
in 2019~2022
As certain products necessitate manual inspection, this process is susceptible to individual subjectivity, variations in training, fatigue, and distractions. These factors contribute to inconsistent inspection standards.
MAIN RESULTS:
Due to the inherent nature of the product properties, all product testing is conducted through destructive methods, making it impractical to test each individual product. However, by gathering sensor data within the production facility, we can leverage machine learning technology to train an AI model.
MAIN RESULTS:
Designing a data model and creating dashboards for Sales, Finance, and Production on Oracle Analytics Cloud has helped project-related personnel save 3-4 days per month.
Create a data pipeline that integrates Manufacturing Execution System (MES), Enterprise Resource Planning (ERP), and sensor data within the production facility. This pipeline will assist production technicians in monitoring products along the production line and preparing for AI applications.