Hire Harvey Y. - Vector Database Engineer

Arizona State DCS Virtual Agent: Built a RAG architecture-based policy consultation chatbot for state government child safety department using Azure AI Search, Azure OpenAI, and Flask framework, integrating voice interaction capabilities. - AWS Intelligent Document Processing: Developed a multi-Agent collaborative medical insurance document processing system using AWS Nova model series and LangGraph workflow engine, achieving end-to-end automation from PDF documents to SQL scripts. - Nashville PD Video Analytics: Developed an intelligent police bodycam video analysis system using AWS Step Functions workflow orchestration, Amazon Rekognition face detection, and OpenCV privacy protection processing with serverless architecture automation. - Elevance IVR Optimization: Optimized pharmaceutical company IVR system using UMAP+HDBSCAN unsupervised learning for new user intent discovery, combined with BERT deep learning classifier and DeepSeek/Llama3.1 large models for intelligent analysis. -Pharma Regulatory Document Generator: Developed an intelligent regulatory document generation system for pharmaceutical company using AutoGen multi-Agent collaborative framework (Creator/Evaluator/Revisor) and Weaviate vector database RAG architecture, successfully deployed to Azure production environment. -Hyatt Sales Automation: Designed an AI sales automation system for international hotel chain using AWS Bedrock Claude 3 and RAG architecture, achieving end-to-end automation from RFP document parsing to sales proposal generation, improving response rate from 5% to 100%. - NYPD Airport Surveillance: Built an airport surveillance intelligent recognition proof-of-concept system using OpenAI GPT-4o-mini multimodal AI and OpenCV video processing, achieving abnormal behavior detection and intelligent scene recognition.

Machine Learning Prediction Analysis: Onboarded creators to livestreaming, leveraging machine learning models to predict potential earnings and design tiered incentive strategies. A/B tests demonstrated a $5M monthly revenue increase compared to the previous approach. - Causal Influence Analysis: Partnered with the operations team to reduce creator churn (attrition). Conducted feature engineering and applied SHAP analysis to identify key drivers (e.g., creators who gained 100K+ followers early had significantly lower churn rates). Designed and A/B tested retention strategies, reducing long-term churn by 20%—now adopted as standard practice. - Heterogeneity Analysis: Collaborated with top influencers (KOLs) to improve retention and spending among young users. Used matching + difference-in-differences (DID) analysis to prove their content boosted engagement. Further UPLIFT modeling revealed the strongest impact on high-value male users, aligning with the company’s target demographic. Findings supported a $10M increase in KOL partnerships. - Experiment Analysis: Designed A/B tests to evaluate business strategies for new users/creators. Monitored real-time data, provided go/no-go recommendations, and measured impact using quasi-experimental methods (synthetic control, matching).