Project 1: LLM-Based Workflow Automation

• Developed and deployed a Large Language Model (LLM) solution to parse unstructured data from emails, images, and tables, incorporating preprocessing pipelines to extract key information, automate database queries, and categorize content
• Implemented an embedding-based Retrieval-Augmented Generation (RAG) pipeline with Maximum Marginal Relevance (MMR) to enhance LLM accuracy by retrieving relevant and diverse historical samples, adapting to varied input patterns
• Leveraged LangChain for scalable LLM application development and LangFuse for performance monitoring

Project 2: Copilot for Business Strategy and Knowledge Management

• Built a LangGraph-based LLM pipeline to extract and organize domain-specific knowledge from presentation slides
• Utilized GraphRAG to structure extracted knowledge, enhancing LLM outputs with relevant, context-specific information
• Designed a standalone knowledge base system with a Streamlit-based UI, allowing business users to manage (insert, delete, update) knowledge, supported by a PostgreSQL database with Pgvector for efficient embedding storage and an MCP server for querying relevant data

Project 3: Operational Optimization

• Developed linear programming models to optimize operational planning processes, improving efficiency through data-driven decision-making
• Maintained automated data preprocessing pipelines to clean and prepare historical data for analytical tasks
• Built evaluation pipelines to assess and refine planning strategies
• Applied stochastic optimization techniques to address uncertainty in operational data, incorporating hard and soft constraints with dynamic penalty mechanisms to model user behavior accurately

• Conducted in-depth research on music generation and music-related classification tasks
• Developed deep learning models for music classification using Stochastic Weight Averaging, mix-up, and gradient clipping, surpassing state-of-the-art benchmarks
• Contributed to the development of content generation

Job duty is mainly on data engineering

• Conduct quality assurance (QA) on mobile applications and web pages to ensure data accuracy and completeness
• Track and process data using various tools such as DBT, Databricks, DBeaver, and Spark
• Query data from Redshift and S3 to extract insights and create reports
• Build interactive Tableau dashboards to visualize and analyze data
• Using Atlassian tools (Jira, Confluence, and BitBucket) for project management / collaboration