Lucan Yan
AI Engineer at DTxPlus- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
-
日本語 Native or bilingual proficiency
-
English Full professional proficiency
-
中国語 Native or bilingual proficiency
Topline Score
Bio
0
/5.0 / Based on 0 ratingsFilter reviews by:
Credentials
-
基本情報技術者
経済産業省Jul, 2021- Sep, 2024
Experience
-
DTxPlus
-
United States
-
Hospitals and Health Care
-
1 - 100 Employee
-
AI Engineer
-
Jun 2023 - Present
● Pioneered the development of a cutting-edge healthcare AI chatbot designed for patient monitoring and digital therapeutics, leveraging natural language understanding and generation techniques (NLU, NLG) to deliver optimal results. ● Trained and leveraged both deep learning and non-DL models using PyTorch for intent detection and slot filling (BERT and XGboost) to implement a robust patient onboarding pipeline. ● Engineered a generative question-answering pipeline employing GPT-3.5 and Llama 2 with in-context learning. The system addresses general user inquiries, enhanced with LangChain and LlamaIndex for persistent memory and proficient information retrieval. ● Built ML and LLM API interfaces in Azure as a service by deploying docker containers to Azure Kubernetes Service (AKS). Show less
-
-
-
WizEats
-
United States
-
Technology, Information and Internet
-
1 - 100 Employee
-
Machine Learning Engineer
-
Mar 2023 - Jul 2023
● Developed a groundbreaking GPT-3.5 powered App Store iOS application to generate recipes in an interactive, step-by-step dialogue with the user, tailored to their needs and inputs. Continually enhance core pipeline and prompt engineering to ensure user satisfaction. ● Implemented a recommendation system that suggests dishes and restaurants based on user’s dietary preferences and current location. ● Engineered backend utilized Django, and Azure Services (Open AI, Cosmos DB, MySQL). Designed and implemented a data pipeline for ETL processes, used Python to extract over 400,000 dishes data observations from the database APIs. ● Utilized bag-of-words and TF-IDF as text-based baselines and substantially improved recommendation outcomes by incorporating product images as inputs. Retrieved similar products by Cosine similarity between image and product description based on CLIP. Show less
-
-
-
Applied Materials
-
United States
-
Semiconductor Manufacturing
-
700 & Above Employee
-
Machine Learning Engineer
-
May 2022 - Aug 2022
Incorporated AI to the world’s largest semiconductor equipment supplier and presented at American Vacuum Society Conference ● Developed and implemented Hadoop-based ETL process to load plasma data using Databricks, and parsed data of different formats into the PostgreSQL database for reporting and analytics, which reduced error data rate by 90% and improved efficiency by 50%. ● Built a deep learning model using LSTM and Transformers for time-series forecasting on 100k+ plasma sequences, utilized TensorFlow and Keras. Improved prediction accuracy (RMSE) by 80% and reduced training time by 70%. ● Optimized the model by conducting a hyper-parameter and architecture search on HPC clusters, used Amazon S3 to store the training data and model checkpoints, enabling easy access to the data and model across team members. ● Built Docker containers and deployed them with AWS SageMaker, which resulted in a 70% increase in efficiency. ● Collaborated with a scientific team to extract data insights and presented them visually using Tableau, revealing multi-dimensional trends and patterns, helping the company to identify key areas of improvement and develop new strategies. Show less
-
-
-
Team AIBOD
-
Japan
-
IT Services and IT Consulting
-
Data Scientist
-
Oct 2020 - Apr 2021
● Developed an E-commerce recommendation system based on 180k+ fashion products using PyTorch and improved the Natural Language Processing (NLP) model. Utilized Google Cloud services such as Dataproc and Cloud storage to enable efficient data processing. Achieved a 150% increase in purchase conversion rate by incorporating both descriptions and images. ● Designed table schemas and constructed datasets of internal company information by web scraping using JavaScript, and HTML, created ER diagrams, and optimized 10+ complex SQL queries for retrieving and manipulating data from a MySQL database. Deployed the database to Amazon RDS for durability. The database improved management team operation efficiency and cost savings by 60%. ● Improved accuracy of electricity consumption forecasting for Toyota by 30% by developing seasonal multi-time sequences prediction algorithms (ARIMA). Optimized the time series model by conducting seasonality analysis and stationary test via ADF test. Show less
-
-
-
Kyushu University
-
Japan
-
Higher Education
-
1 - 100 Employee
-
Research Assistant
-
Mar 2019 - Mar 2021
● Collaborated with researchers to reveal the mechanics of cell division which has relevance to tumors and visualized data insights. ● Conducted data labeling and augmentation by calculating Fourie Coefficients of cell images, and built supervised ML models (LDA, Naive Bayes) using MATLAB to classify 4000+ cell phenotypes simultaneously during cell division. ● Improved data processing efficiency by 80% and achieved a 5% classification error; leading to a paper published in Top Journal (PNAS). Show less
-
-
Education
-
University of Pennsylvania
Master's degree, Computer Science & Machine Learning -
Kyushu University
Master's degree, Machine learning & Computational Biology -
Kyushu University
Bachelor's, Physics