Jinghan Huang

About Jinghan Huang

I specialize in building AI-driven products that solve real-world problems efficiently. My work spans natural language processing, computer vision, recommendation systems, and predictive analytics. I focus on creating practical, scalable, and user-centered AI systems across industries. AI Application Areas: Natural Language Processing: Chatbots, language understanding, summarization, sentiment analysis, and information retrieval. Computer Vision: Image classification, object detection, facial recognition, and OCR. Recommendation Systems: Personalized ranking models for content, commerce, and real estate platforms. Predictive Modeling: Demand forecasting, behavior prediction, churn modeling, and risk assessment. Autonomous Systems: Agent-based decision systems, reinforcement learning applications, and automation tools. Cross-Industry Use Cases: Healthcare: Diagnostic models, drug discovery, patient triage Finance: Fraud detection, credit scoring, algorithmic trading Real Estate: Dynamic pricing, AI search and matching (e.g, Estay.ai) E-commerce: Personalization engines, inventory optimization Education: Adaptive learning platforms, automated grading I believe in applying AI responsibly, with a focus on transparency, fairness, and real impact.

Experience

1. Lead Ai Engineer

   Propbotics

   May 2023 — Present · NY, US

   Spearheaded the full-stack development of Estay.ai, an AI-driven rental platform that matches users to NYC apartments via natural language chat; scaled to users in 6 months.• Built a multi-agent system leveraging LLMs (OpenAI, DeepSeek) and FAISS-based retrieval, enabling dynamic conversation workflows for apartment matching, reasoning, and FAQ resolution.• Optimized data processing by developing a robust LLM orchestration framework leveraging Pixtral OCR to extract data from large PDFs, enabling effective training and testing dataset creation, which guaranteed 95% accuracy in evaluations.• Developed an LLM orchestration system to verify claims based on multiple documents, resulting in a 27% increase in accuracy and creating slides to present key findings from research papers.• Designed and deployed retrieval-augmented generation (RAG) pipelines, significantly improving answer accuracy and context relevance by 57%, boosting user retention and lead conversion by 5 times.• Developed a self-reflective reasoning agent that analyzes failed search outcomes, identifying causes and guides users to adjust criteria through multi-turn dialog, resulting in a 150% increase in user retention for failed-to-successful conversion flows.• Architected and implemented a MySQL-based structured data backend from scratch, including schema design, ETL pipelines, and full integration with LLM agents to enable structured reasoning over 450+ apartment buildings.• Developed apartment ranking algorithms using Google reviews, listing metadata, and user preferences to enhance recommendation precision and trust, boosting top-3 recommendation click-through rate by 40% and improving perceived quality in user feedback

Education

• Columbia University

  Master of Arts - MA, Statistics

• University of Illinois Urbana-Champaign

  Bachelor of Science - BS, Mathematics