Wei Yang

Gain hands-on experience on software development and data pipeline implementation

• AI Chatbot & RAG Systems:
  • Led full-cycle development of a production-grade RAG conversational AI chatbot using Azure OpenAI, Redis, and Weaviate vector database; architected sticky session routing and load balancing to support 1K+ DAUs
  • Designed multi-turn conversational memory with Redis persistence and context window management; built agentic workflows with function calling to orchestrate custom tools and external APIs
  • Implemented security guardrails including input sanitization, content filtering, and PII masking to ensure safe and compliant AI interactions
  • Delivered real-time AI Overview and autosuggest powered by live user queries with caching layer for low-latency responses; built scheduled report pipelines and user feedback loops to continuously improve relevance and quality
  • Engineered scheduled multilingual (I18N) indexing pipelines across 10+ heterogeneous data sources, combining keyword-based and semantic hybrid search with semantic reranking, multi-channel query routing, query expansion, and iterative retrieval
• Infrastructure & Platform Engineering:
  • Architected scalable full-stack infrastructure: VM provisioning, runtime orchestration, and offline pipelines for knowledge base synchronization and cache optimization
  • Maintained high reliability and low-latency performance through proactive monitoring and tuning
  • Applied data-driven insights to evolve CMS architecture and scale web platforms
• Data Integration & Search Optimization:
  • Designed 7+ MuleSoft API integration workflows for incremental delta updates of partner accounts and locations
  • Integrated multi-source data into a Solr and Elasticsearch–based search layer with cache optimization
• Personalization & Machine Learning Pipelines:
  • Developed User-to-Item and Item-to-Item recommendation pipelines with rolling cache and Akamai CDN integration
  • Built offline ML pipelines to optimize sales funnel conversion and marketing campaign targeting
• Content Platform (AEM):
  • Developed licensable software modules on Adobe Experience Manager to streamline content authoring and multi-channel publishing

Language features pattern detection with Deep Learning models, data parsing and statistics modeling

• Cleaned and structured 94,581 earnings call transcripts to explore 96 language factors influencing stock return
• Achieved 72% accuracy on text classification with domain-adapted BERT language model using Pytorch
• Conducted text analysis, sentiment analysis, and text mining utilizing NLP techniques with SpaCy and NLTK
• Set up and managed a remote GPU environment on Ubuntu for Machine Learning and Deep Learning tasks
• Developed tutorials on implementing deep learning models using related Python packages

Develop a web application for furniture and rental information exchange

• Led team of four students to design and develop a web application with Node.js and Express
• Designed document schema on MongoDB and wrapped CRUD operations as RESTful APIs
• Implemented Login and user-specific functions such as account signup and authentication system
• Developed features for rental and furniture information exchange such as comments, search, dashboard

Develop a web application for furniture and rental information exchange

• Created and implemented recommender engine with users, items, and interaction records from JD.com
• Integrated multiple memory-based and model-based collaborative filtering algorithms to make recommendations
• Simulated on 7,000 pre-defined user-item interaction samples and attained 77% Top-10 Accuracy

Develop a mathematical model based on public data and metrics to measure and predict the vibrancy of the city in the U.S.

• Constructed a vibrancy index to interpret and predict the prosperity trend of cities in the U.S.
• Explored, collected, and blended data from Google POIs, Instagram, Zillow, Bureau of Labor Statistics with Pandas
• Performed feature engineering on panel data, and fine-tuned models for prediction with XGBoost and Keras

Data scraping, Parsing and information extraction, topic modeling with Clustering algorithms and Neural Networks

• Scraped, cleaned and structured Job Descriptions textual data from INDEED.com
• Filtered noisy data by selecting clusters from K-means algorithm and visualized skillsets distribution
• Categorized, analyzed skillsets trend after Topic Modeling and Aspect Extraction Deep Neural Network

Feature engineering, Data analysis and modeling

• Performed data cleaning and sentiment analysis for over 20 million related Tweets in second granularity
• Delivered feature engineering for financial indicators like MACD and RSI etc
• Analyzed, finetuned, and back-tested for over five types of deep learning model to increase the portfolio return