JiunYi Yang (楊鈞宜)

Senior AI Software Engineer / Agent Framework Expert

Senior AI Engineering Leader with 6+ years of experience architecting enterprise-grade RAG frameworks and agentic AI solutions for industry leaders. Published researcher with TCNNet-9B in Nature Scientific Reports. Team leader with proven track record of mentoring engineering teams and delivering high-performance AI systems achieving 95%+ accuracy.

Key Achievements

95%+

FAQ Response Accuracy

Advanced multi-stage RAG framework achieving breakthrough accuracy

30K+

Daily Conversations

Scalable AI systems handling massive conversation volume

3x

Faster Deployment

Agent Scenario Protocol enabling rapid client customization

<5s

Response Time

High-performance systems with sub-5-second latency

Professional Experience

Senior AI Software Engineer

RaccoonAI • Taipei, Taiwan

Aug 2024 - Present

Omnichannel AI CRM solution integrated into platforms and social channels (Line, FB, IG)

First/only AI hire pre-funding - Shipped MVP in 1 month, drove customer customizations. Company raised AppWorks Pre-A round funding 10 months after joining.

Led cross-functional AI engineering team of 3, mentoring 2 new engineers in enterprise-grade chatbot development - Raccoon AI Chat.

Achieved 95%+ accuracy in FAQ responses, handling 70% of customer inquiries autonomously through advanced multi-stage RAG with integrated BM25 retrieval.

Architected Agent Scenario Protocol enabling modularized agent configuration, increasing client PoC and deployment speed by 3x.

Notable Clients:

BenQ
WeMo
Rhinoshield
Positive Grid
SWAG
Lianli
VERVE
BitoPro
Uniswap
Farglory

Founder & Freelancer AI Developer

DataAgent • Taipei, Taiwan

Jun 2023 - Aug 2024

AI technology consulting, AI product development, and content monetization

Achieved 40% increase in client engagement and 25% improvement in marketing ROI through pioneering RAG applications for marketing strategy.

Created TaiwanEcoChat-9B, an LLM tailored for ESG content, surpassing original baseline on Taiwanese ESG knowledge benchmark.

Architected chat-based e-commerce product recommendation system with GraphRAG, improving accuracy by 30% and reducing user exploring time by 50%.

Developed TCNNet-9B, a domain-adapted LLM for networking & cybersecurity, surpassing baseline in Taiwanese networking knowledge benchmark.

NLP Engineer

Cathay Financial Holdings • Taipei, Taiwan

Jul 2021 - Oct 2023

Major financial conglomerate actively promoting AI and data applications integration

Achieved 93% F-score in financial information extraction models, reducing data preprocessing time by 40%.

Enhanced financial forecast accuracy by 30% with relationship classification model for investment research dashboard.

Achieved 240% increase in precision for Anti-Money Laundering (AML) using LSTM/GRU for abnormal user predictions.

Data Analysis Engineer

Tagtoo AD • Taipei, Taiwan

Jul 2019 - Jun 2020

Ad tech startup specializing in high-conversion performance advertising

Achieved 19% increase in new registrations and won 3rd place of Digital Singularity Award with cross-site user clustering using K-Means.

Reduced cost per lead by 31% and increased ad conversion rates by 22% with data-driven advertising strategies.

Publications & Research

Localized Large Language Model TCNNet 9B for Taiwanese Networking and Cybersecurity

Nature Scientific Reports 15, 8418 (2025) • Published March 20, 2025

LLM
Cybersecurity
Traditional Chinese
Domain Adaptation
  • Achieved 2.35x improvement in Q&A task accuracy over baseline models
  • 37.6% increase in domain expertise comprehension through specialized development
  • Enhanced product recommendation relevance by 29.5% using advanced fine-tuning

Eco Smart Guide: An LLM and RAG-based Platform for Streamlined ESG Information Access

Best Conference Paper Award

RAG
ESG
Information Retrieval
Best Paper

Differencing Time Series as Feature Extraction for Intradialytic Hypotension Prediction

2021 IEEE 3rd Eurasia Conference on Biomedical Engineering • Best Conference Paper Award

Machine Learning
Healthcare
Time Series
Best Paper
View Publication

Education

Master of Management Information Systems

National Cheng Chi University • Taipei, Taiwan

Sep 2019 - Jun 2021

GPA: 4.2/5.0 (Top 8% of class)

Bachelor of Industrial Engineering and Management

National Chiao Tung University • Taipei, Taiwan

Sep 2013 - Jun 2017

Technical Expertise

Expert Level

Retriever-Augmented Generation (RAG)
Large Language Model Architecture
Prompt Engineering
Python (AI/ML)
Vector Databases (Qdrant, pgvector)
API Design & Microservices
Cloud Services (GCP/AWS/Azure)
Enterprise AI Architecture

Experienced

Docker & Kubernetes
GraphRAG
Fine-tuning & PEFT
LangChain
FastAPI
PostgreSQL
Redis
CI/CD

Open Source Contributions

A-baoYang/alpaca-7b-chinese

A LLaMA-7B instruction fine-tuning project using Traditional Chinese datasets. Responsible for providing a runnable fine-tuning process in the early stage of LLM emerging.

View on GitHub

A-baoYang/LLM-Finetune-Guide

An instruction fine-tuning guiding project provides LoRA and P-Tuning methods with easy-running scripts for practitioners.

View on GitHub

Speaking & Community Engagement

Neo4j Dev Conference #NODES 2024

"Enhancing Chat-Based Customer Recommendations With GraphRAG and Hybrid Search"

Oct 2024

DataCon TW 2022

"Financial Knowledge Graph & Investment Applications"

Oct 2022
Watch Talk

PyCon TW 2021

"Python Data Science Application in Intradialytic Hypotension Prediction"

Sep 2021

Tech Blog Creator & Social Media Influencer

Founder and main author of DataAgent, a popular blog focused on AI, data science, and technology trends. Regularly write articles that simplify complex technological concepts for a broad audience.

Impact: Top posts reaching over 2,000 shares and 700 likes on Facebook, highlighting strong capability in creating impactful content that resonates with tech professionals and enthusiasts.