YuehChun (Hugo)

YuehChun (Hugo)

Senior AI Engineer & MLOps Specialist

↓ Scroll

A believer in data.
55% Scientist. 25% Analyst. 10% Engineer. 10% Dreamer.

Senior Machine Learning Engineer & MLOps Specialist with 10+ years of ML experience. Currently building production-ready AI agent systems with LLM and multi-agent orchestration at Mal Bank, UAE. Previously focused on reinforcement learning for quant-trading, MLOps platform architecture, and large-scale distributed training at Polymer Capital Management.

Interested in researching and developing:

→ LLM Applications & Agentic AI Systems

→ MLOps & ML Infrastructure (Kubernetes, Ray, Distributed Training)

→ Reinforcement Learning for Quantitative Trading

→ Real-world AI/ML Applications (Finance, Manufacturing, NLP, Computer Vision)

Where I've Made Impact

Senior AI Engineer

Mal Bank · Abu Dhabi, UAE

2025 — Present

• Built Personal Secretary Agent with ReAct framework and Supervisor pattern for multi-agent orchestration

• Integrated memory functionality, GraphDB for user profiles, and RAG for tool selection

• Architected gRPC-based microservices infrastructure with Protocol Buffers, supporting high concurrency and low latency

ML Researcher (MLOps & Quant Infrastructure)

Polymer Capital Management · Taipei

2023 — 2025

• Developed RL (PPO) strategy enhancement for futures trading, achieving 2x RoMaD

• Built Variational Autoencoder for noise reduction and synthetic market data generation

• Created knowledge-based LLM tool using OpenAI API, cutting daily report preparation by 1+ hour

• Architected end-to-end MLOps platform with Ray, scaling RL workloads to 1k GPU nodes, reducing training time by 90%

Data Scientist — AI Engineer

ChungHwa Telecom · Taipei

2016 — 2023

• Applied adversarial inverse RL to energy control policy optimization, achieving 10% energy consumption reduction

• Deployed edge anomaly detection (Inception V3) on Jetson Nano with MQTT + AWS Greengrass

• Improved CTR by 50% through refined feature selection and enhanced label granularity in telecom AdTech

• Optimized production lines in manufacturing, reducing MAE by more than 55%

• Engineered high-throughput Spark/HiveQL ELT pipeline for TB-scale logs, reducing latency to under 10 minutes

• Built automated analysis applications for COVID-19 contact tracing, traffic analysis, and dynamic reporting

Intern

Industrial Technology Research Institute

2015

• Coded in Ruby on Rails, accessing web APIs from Facebook, Google, Dropbox, Nexmo

• Learned collaboration workflows using Git version control

Earlier Roles

Part-time MIS & Associate Programmer

2009 — 2013

• Part-time MIS at NUTC-CSIE: managed department server, established support systems (project exhibition, voting service)

• Associate Programmer at 嵐宇科技: PHP development, database schema design, customer communication

• Associate Programmer at 網通數位科技: built front-to-end web services independently, learned MySQL, PHP, JavaScript, HTML, CSS

What I've Built

LLM & GenAI

Personal Secretary Agent

Multi-agent orchestration with ReAct framework, memory, GraphDB, and RAG

Production-ready AI agent system at Mal Bank with Supervisor pattern for multi-agent coordination. Integrated memory functionality and GraphDB for user profiles.

MLOps

Ray/MLOps Platform

End-to-end ML platform scaling RL workloads to 1k GPU nodes, reducing training time by 90%

Reinforcement Learning

PPO Trading Strategy

RL-based futures trading with custom reward shaping, achieving 2x RoMaD

Deep Learning

Variational Autoencoder

Noise reduction + unlimited synthetic market data generation for backtesting

Data Engineering

TB-Scale ELT Pipeline

High-throughput Spark/HiveQL pipeline reducing ingestion-to-reporting latency to under 10 min

Deep Learning

Edge Anomaly Detection

Inception V3 on Jetson Nano with MQTT + AWS Greengrass IoT automated deployment

LLM & GenAI

Knowledge-based LLM Tool

OpenAI-powered report aggregation cutting daily report prep time by 1+ hour

Machine Learning

AdTech CTR Optimization

Improved click-through rate by 50% via refined feature selection and label granularity

Technical Expertise

Programming

Python SQL TypeScript Terraform Java PHP

ML / AI

LLM & GenAI Transformer RL (PPO) GraphRAG XGBoost LSTM VAE

Infrastructure

Kubernetes Ray & Spark AWS gRPC HiveSQL

DevOps

MLflow & DVC Neo4j Docker & CI/CD FastAPI CDK8s

Academic Contributions

基於社群媒體的代表性新聞提取方法:以 3D 列印為例 — CCBDA 2016
Publication
Mining Web Browsing Behavior on Tourism Sites in Taiwan — ACEAT 2014
Publication
關聯式規則平行化效能比較之研究 — IMP 2013
Publication
MapReduce 架構下循序樣本探勘演算法之效能分析 — TANET 2013
Publication
102 學年度台中市大學模範生
Award · 2014
103 學年度大專優秀青年
Award · 2014

Academic Journey

M.S. Computer Science & Engineering

National Chung Hsing University · 2014 — 2016

B.S. Computer Science & Information Engineering

National Taichung University of Science and Technology · 2010 — 2014