initializing...

KING WAI SIU

build()

I build AI products that turn complex data into clear, useful decisions. Right now I’m co-building HKLawyer.ai, using RAG systems and LLMs to open up Hong Kong legal knowledge to regular people. When I’m not training models or digging through data, I’m obsessed with making experiences that don’t suck.

View Projects

About

Hey! I'm King Wai (or just Toby). I studied New Media thinking I'd make contemporary art, but fell into Data Science and discovered I'm passionate about turning data into actionable insights that solve real problems.

# quick_about.py
class KingWaiSiu:
def __init__(self):
self.role = "AI Engineer & Data Analyst"
self.location = ["HH", "HK"]
self.expertise = [
"Data Analysis",
"LLM Integration",
"RAG Systems",
"Full-Stack AI"
]
self.current = "HKLawyer.ai"
def build_cool_stuff(self):
return "Always."
visitor@portfolio:~$ cat philosophy.txt
"Good design is obvious. Great design is transparent."
— Me, probably inspired by someone smarter
visitor@portfolio:~$ █

Experience

Building things that matter, one commit at a time

Associate AI Engineer

DHL Data & AI GmbH

Bonn, Germany
05.2026 - Present
GenAI OCR LLM Prompt Engineering Analytics Gemini Pydantic Next.js Power BI
  • Engineered data ingestion and processing pipelines for DHL’s Document AI, integrating Gemini through the enterprise GenAI Gateway and using rasterized pages to improve document extraction
  • Built Grafana dashboards to monitor API costs, token usage, and errors across AI services, supporting cost optimization and operational troubleshooting
  • Developed comprehensive Power BI dashboards to monitor conversational analytics, automated resolution rates, and user-flow performance for the Deutsche Post & Parcel AI voice bot (Jana)
  • Built a Pydantic AI multi-agent system for DHL’s internal-control validation, with Python orchestration, deterministic compliance checks, and agent tool calling for approved rule and template retrieval
  • Developed a GenAI React-based web application for summarizing feedback, including an interactive frontend and structured AI-generated recommendations for DHL Consulting
  • Designed the backend with MongoDB and RabbitMQ for the DHL Express Customs AI Agent

Data Analyst

ALTEN Technology GmbH

Technical Direction
08.2021 - 04.2026
SQL Python Azure Fabric Power BI Power Automate Power Apps UX/UI Design
  • End-to-end development of interactive Power BI dashboards and semantic models (schema design, custom DAX, and Power Query M) from scratch in close collaboration with business stakeholders across project management, HR, finance, and IT, including migrating legacy Excel databases and reports into centralized SharePoint and Power BI architectures
  • Building and maintaining automated ETL/ELT pipelines using SQL, Microsoft Fabric Dataflows, and Databricks, integrating enterprise data sources (SAP and SharePoint), and managing Dev/Production workspaces, service accounts, and the Power BI Gateway for scheduled data refreshes
  • Enforcing data governance and GDPR compliance by removing personally identifiable information (PII) during Power Query transformations and configuring Row-Level Security (RLS) within analytical semantic models
  • Performing ad-hoc data extraction, preprocessing, and custom visualizations using Python and SQL, alongside analyzing operational KPIs and user behavior to deliver actionable insights for project directors and senior management
  • Developing and containerizing (Docker) full-stack GenAI applications, technical RAG chatbots (leveraging LLM-based OCR, document-level chunking, embedding models, and vector databases), and LLM automation workflows using Python, TypeScript, Azure AI Foundry, and Copilot Studio
  • Engineering custom Microsoft Power Platform solutions (Power Apps and Power Automate) using Power Fx and REST API integrations (such as automated Microsoft Teams provisioning) to modernize legacy business processes
  • Leading cross-functional solution rollouts, including requirements gathering, coordinating User Acceptance Testing (UAT), writing technical documentation, and conducting end-user onboarding tutorials

Working Student

BASF Coatings GmbH

Procurement Controlling & Intercompany Business
01.2021 - 05.2021
SQL Power BI Power Apps
  • Managed ETL of data from various sources
  • Created SQL views in Azure SQL database

Intern

BASF East Asia Regional Headquarters Ltd.

Global Controlling, Dispersions & Pigments
05.2020 - 10.2020
Python Azure Databricks SQL Power BI
  • Transformed data from Azure Data Lake with Spark in Databricks
  • Built a chemical supplier recommendation system using collaborative filtering
  • Developed linear regression model using Python for sales forecasting
  • Created Power BI dashboards with data from SAP BW and Azure Data Lake

Intern

BASF East Asia Regional Headquarters Ltd.

IT Solutions, Advanced Business Analytics
10.2019 - 03.2020
Python NLP FastAPI Docker Azure
  • Built HR recruitment tool using NLP and TF-IDF
  • Deployed Python backend with FastAPI and Docker

Projects

Things I've built that I'm actually proud of

HKLawyer.ai

Next.js LLM Supabase RAG LangChain

AI legal platform making free legal advice accessible. Multi-embedding RAG system trained on 20+ years of Hong Kong case law. Autonomous content generation and marketing agents.

  • Build a comprehensive legal AI chatbot with Next.js & Supabase (Postgres)
  • Implement multi-embedding RAG with 20+ years of HK Judiciary data
  • Develop AI-powered SEO automation system using LangChain and RAG
  • Create autonomous content generation and marketing agents
12K+ Questions
52K+ Visitors
1M+ Impressions
Visit Site

Deepstory

PyTorch GPT-2 Flask

Deepstory is an AI approach to storytelling. It is a Flask-based LLM web application that uses several PyTorch Deep Learning models as the backend:

  • Generate story text with a fine-tuned GPT-2 model using The Witcher novel text data
  • Synthesize the text into speech using a custom-trained PyTorch based DCTTS model using manually scraped character voice data + transcript from The Witcher 3 game
  • Animate it into a video based on the character's image using Speech-Driven Animation & First Order Motion model
98 GitHub Stars
View on GitHub

Education

MSc in Management & Data Science

Leuphana Universität Lüneburg

10.2026
Data Science Machine Learning Deep Learning Artificial Intelligence Large Language Models RAG Systems Agentic Systems Computer Science Statistics Matlab Management
  • Analyze massive and complex datasets, design statistical models
  • Practical experience in Python, PyTorch, numpy, Pandas, and Git
  • Modules such as Probabilistic Modeling, Optimization Techniques

BAS in New Media

City University of Hong Kong

07.2020
Computer Science Comtemporary Arts Generative Art Graphics Design Game Development Animation UI/UX Design HCI
  • Dean's List Semester B 2019/2020
  • Dean's List Semester A 2019/2020
  • Dean's List Semester A 2015/2016

BA in Digital Media

Leuphana Universität Lüneburg

07.2020
Media Theory Media Production Media Design Cultural Studies Computer Science Software development Arduino
  • Minor in Popular Music Studies

Skills

Programming

Python Rust C++ JavaScript TypeScript Git VS Code Cursor

AI & Data Science

PyTorch Transformers LangChain RAG Systems scikit-learn pandas NumPy

Cloud & DevOps

Azure AWS GCP Docker Kubernetes Databricks Fabric

Full-Stack Dev

React Next.js Vue.js FastAPI Node.js Django PostgreSQL RESTful APIs

Analytics

Fabric Power BI DAX Tableau Spark SQL ETL/ELT

Design & UX

Figma Canva Sketch Photoshop Illustrator Premiere Pro After Effects User Research

Certificates

GH-600

GitHub Agentic AI Developer

PL-300

Power BI Data Analyst Associate

DP-600

Fabric Analytics Engineer Associate

DP-700

Fabric Data Engineer Associate

AI-900

Azure AI Fundamentals

IBM Data Science

Professional Certificate

Let's Connect

# contact.py
def reach_out():
email = "thetobysiu@gmail.com"
location = "Bonn / Hong Kong"
available = True
interests = [
"AI Projects",
"Collaborations",
"Coffee Chats",
"Opportunities"
]
if you.have_cool_idea:
send_email(to=email)
return email