AI Engineer & ML Specialist

Vincent Hsia

Software Engineer / AI Engineer

Hi, I'm Vincent Hsia!

I am a Master's student in Management Information Systems at National Chengchi University, specializing in LLM System Engineering and Generative AI applications.

With a focus on moving AI from prototype to production, I have built and deployed agentic workflows and RAG-driven pipelines for document understanding and automated regulatory analysis. My expertise lies in bridging the gap between large language models and robust backend systems—integrating vector databases and verification loops to ensure system reliability and high-quality decision-making.

Build with AI Ship end-to-end systems Write technical notes

GitHub LinkedIn 113356046@g.nccu.edu.tw

Research Interests

GenAI Application ( MCP, Agent, RAG )
Neuro-Symbolic System
Data Engineering
LLM Deployment
Golden Retriever 🦮

Skills

* Core AI & LLM Engineering:
* LLM Frameworks: AutoGen (Multi-Agent), LangChain, LlamaIndex, LiteLLM Proxy
* GenAI Paradigms: Agentic RAG (CRAG, Reranking), Agent-based Workflows, Prompt Engineering, LLM-as-a-Judge
* Model Optimization: Fine-tuning (LoRA), Model Context Protocol (MCP), Formal Verification (SMT Solvers)
* AI Frameworks: PyTorch, TensorFlow
* Software Engineering & Web:
* Languages: Python (Expert), JavaScript, C++
* Backend: FastAPI, Django, Flask, Jinja2, RESTful APIs
* Frontend: React.js, Angular.js, Chainlit UI (AI-Native Interface)
* Data Engineering: Apache Airflow, Multimodal ETL Pipelines, Data Ingestion/Extraction
* Database & Infrastructure:
* Vector/Graph DB: ChromaDB, Neo4j
* SQL/NoSQL: PostgreSQL, MySQL, MSSQL, MongoDB
* DevOps & Tools: Docker, Git, GitHub Actions (CI/CD), Make, GCP (Google Cloud Platform)

Publications

KDD 2025 ML in Finance Workshop 2025

A Hybrid Framework for Financial Regulatory Compliance: Integrating LLMs and SMT Solvers for Automated Legal Analysis

Yung Shen Hsia, Fang Yu

2nd ACM International Conference on AI-powered Software 2025

Neuro-Symbolic Compliance: Integrating LLMs and SMT Solver for Automated Financial Legal Analysis

Yung Shen Hsia, Fang Yu, Jie-Hong Roland Jiang

Projects

2025/07 - 2026/06

Carbon Credit Assessment and Compliance Verification : Exploring Carbon Footprint and Cost-Effectiveness Optimization Strategies

Developed an autonomous Multi-Agent system using AutoGen and SMT Solvers to formalize carbon-credit regulations and optimize trading strategies with self-correcting logic.

2025/01 - 2025/12

The Application of AI Agent Platform to the Study of Producing Investment Analysis Reports

Architected an interactive Multi-Agent system using SMT Solvers and Chainlit to transform financial regulations into logical constraints, enabling automated compliance optimization with human-in-the-loop flexibility.

2023/07 - 2023/09

Medical QA Optimization: Fine-Tuning LLaMA-7B via LoRA on PubMed Datasets

Fine-tuned the LLaMA-7B model using LoRA on the PubMedQ&A dataset. We achieved 68% accuracy and 51% F1 score during evaluation.