An innovator blending the narrative power of contemporary art curation with modern AI architecture. I bridge the gap between complex engineering paradigms (LangGraph Multi-Agents, custom RAG pipelines, prompt optimization) and high-fidelity aesthetic delivery (immersive interactive installations, digital media, location-based experiences).
As a former Exhibition Director (managing multi-million dollar international projects) and a Digital Arts MFA, I build AI systems that aren't just functionally robust, but conceptually deep and aesthetically premium.
- Why this matters: In a world of generic API wrappers, I design systems with high user empathy, rigorous structure, and tailored intelligence—whether it's an AI-driven dialogue system representing historical artists or an automated investment and research laboratory.
|
|
|
Below are selected personal and professional engineering projects showing the practical execution of my technical toolbox.
Trading Agents (A-Share Optimized) LangGraph Multi-Agent State-Machine
Multi-Agent Financial Debate & Quantitative Decision Engine
- The Core: Leverages LangGraph to coordinate a multi-agent assembly (Analyst, Risk Officer, and Portfolio Manager) that debates quantitative signals and market states before executing simulated portfolio rebalancing.
- Why it shines: Displays mastery of complex state machines, cyclical graphs, human-in-the-loop gates, and multi-agent coordination—the absolute state-of-the-art in modern AI engineering.
wiki-agent LangChain Advanced RAG Obsidian Graph
Automated Cross-Domain Knowledge Ingestion & RAG Indexing Platform
- The Core: A zero-cost, production-grade pipeline that automatically ingests complex documents (PDF, DOCX, PPTX), extracts structured data using OCR and LLMs, auto-generates bidirectional Obsidian knowledge graphs, and pre-loads
bge-m3semantic embeddings into a persistent Chroma DB server. - Why it shines: Implements multi-provider LLM intelligent routing with circuit breakers, directory watching, idempotent vector sync, and an active FastAPI query service layer.
model-scout Next.js FastAPI Observability
LLM Performance & Synthetics Observability Dashboard
- The Core: Runs automated synthetic transactions against various LLM endpoints to evaluate real-time latency, API drift, availability, and semantic consistency.
- Why it shines: Highlights full-stack architecture, combining elegant reactive Next.js frontends with high-throughput FastAPI backends for technical observability.
quant-lab Python AkShare LLM News Filter
A-Share News Extraction & AI Sentiment Evaluator
- The Core: Ingests high-frequency financial feeds using AkShare, filters market noise through a custom 3-engine data pipeline, and executes LLM semantic scoring to gauge market momentum.
- Why it shines: Demonstrates clean integration of industrial data scraping, real-world finance API handling, and LLM classification metrics.
Smoothie Reader FastAPI Edge-TTS Neural TTS
AI-Powered Neural E-Reader with Real-Time Translation
- The Core: Converts static EPUB files into a highly dynamic neural audio experience with real-time translation layers, localized speech synthesizers, and adaptive UX.
- Why it shines: Beautifully merges media APIs, TTS conversion engines, and frontend design paradigms.
sugar-bee Flask SQLite Health Tech
Personal Blood Glucose Monitoring & Analytics Application
- The Core: A secure, database-backed tracker designed to log, model, and graph daily blood sugar levels, tracking metrics for metabolic feedback.
- Why it shines: Proves execution of reliable database schemas, secure session management, and standard MVC web application fundamentals.
Auto Nutrition Python MyFitnessPal API USDA API
Automated Calorie & Dietary Macro-Nutrient Sync Pipeline
- The Core: Integrates USDA food databases and MyFitnessPal APIs to automatically clean, calculate, and log daily metabolic and dietary intake.
- Why it shines: Proves capability in third-party API OAuth/handling, data normalizations, and headless automation.
🌌 UNKNOWN | YIN XIUZHEN Exhibition (Executive Curator & Tech Lead)
An internationally recognized interactive installation spanning three stories at ZHI Art Museum.
- Tech Implementation: Integrated real-time API global news streams, captured visitor facial expressions (sentiment/emotion mapping), and fed the data into a custom AI visual generator displayed on a custom 4-meter LED ring.
- Physical Transformation: Aggregated daily emotional metrics to automatically trigger unique high-precision 3D-printed physical celestial models, turning fleeting human emotions into permanent museum artifacts.
🗣️ Artist Dialogue Prototype (AI-Powered Art Research RAG)
An interactive prototype utilizing Retrieval-Augmented Generation (RAG) to breathe life into cultural archives.
- Tech Implementation: Ingested hundreds of pages of historical catalogs, letters, and art critiques into a structured knowledge base. Optimized multi-turn conversational agents to simulate historically-grounded dialogues with deceased artists.
🎨 "Your Cubist Portrait" (Interactive Generative Experience)
A location-based interactive installation for museum visitors.
- Tech Implementation: Coordinated custom neural image-style transfer pipelines. Visitors uploaded portraits via mobile terminals, which were stylized into Cubist masterpieces on-the-fly and outputted to a high-speed dye-sublimation printer.
- Alibaba Cloud ACP (Alibaba Cloud Certified Professional) - Large Model Senior Engineer (高级大模型算法工程师)
- AI Research Tracker: Maintained a systematic AI Literature Reading Ledger tracking 140+ core papers across Multi-Agent systems, prompt tuning, and cognitive architectures.
- MFA in Digital Arts - Bowling Green State University, USA (GPA: 3.96 / Full Tuition Scholarship)
- Nominee for the Distinguished Studio Arts Award & Member of the Golden Key International Honour Society
- BA in Digital & Multimedia - Shandong University of Arts (National Scholarship Recipient, Outstanding Graduate)
- 📧 Email: yhn0535@gmail.com
- 📱 Phone: +86 13162006535
- 🗺️ Target Locations: Shanghai / Suzhou (China) & Jakarta / Bali (Indonesia) | Open to remote, hybrid, or on-site roles
- 🌐 Portfolio: Ask me for my comprehensive 55-page interactive PDF portfolio!