M.S. in Learning Sciences / R&D Engineer

07/2022 ~ 07/2024

Text Search Algorithms
Webhook API Development
AI Applications
National Tsing Hua University (Institute of Learning Sciences and Technologies) logo

National Tsing Hua University (Institute of Learning Sciences and Technologies)

Hsinchu

Published two papers (including Best Paper Award), researching LLM prompting strategies and self-directed learning resource recommendation systems, utilizing web scraping, Milvus vector database, Flask API, and RabbitMQ to achieve efficient data retrieval and concurrent processing, optimizing system response time from 1 minute to 50 milliseconds.
Trello Knowledge Assistant 2.0

Trello Knowledge Assistant 2.0

  • Used SBERT (Sentence-BERT) for text vectorization combined with Milvus vector database, reducing resource recommendation computation time (1 min → 3 sec) while maintaining search accuracy.
  • Built WebAPIs with Flask and integrated Trello Webhook, reducing system response time from 1 minute to 50 milliseconds after restructuring the data flow.
  • Fuzzy Text Search Algorithm
  • Converted text into dense vector representations via SBERT for precise semantic similarity computation and text matching.
  • Maintained and optimized the Milvus vector database to ensure stability and performance of large-scale vector retrieval.
  • Combined NLP techniques with search algorithms to improve search result accuracy and system user experience.
  • Flask API Development
  • Designed RESTful APIs based on the Flask framework, providing stable and flexible backend services.
  • Optimized code architecture following best practices to ensure API scalability and maintainability.
  • Wrote unit tests and technical documentation to ensure proper system handover.
查看專案