Wen-Kai Chung
I Am Passionate
An enthusiast for building innovative ideas through the synergistic blend of mathematics, economics, and programming expertise.
ABOUT ME
Success is walking from failure to failure with no loss of enthusiasm. — Winston Churchill
I am Wen-Kai Chung
A professional data scientist specialized in time-series and NLP
I develop models with robust object-oriented-programming (OOP) principles, monitor them with MLOps mindset, and deploy them in cloud environment. Employers and clients are happy about my working results because they can achieve not only better model performance but also the efficient model versioning and the balance between performance and stability. Checkout my projects and their feedbacks here!
Data science is the field I thoroughly enjoy, where we extract valuable insights and enhance our clients' lives through data-driven automation and decision-making. Download my CV below to see my ability in time-series, NLP, and causal inference!
Germany
wkaichungtw@gmail.com
(+49) 152 2215 7727
4+ years
3+ years
1 year
My Skills
-
Python
-
Matlab
-
LaTeX
-
C++
-
R
-
SQL
-
HTML
-
JavaScript
Experience & Education
-
Data Science Intern & Thesis Student
Lufthansa Technik- Implemented Hierarchical Flight Hours Forecast model, beating company's existing model and getting lots of business attention.
-
Data Science Intern
Forecasty.AI- Led the beta launch of the AutoML platform.
- Deliver a model that successfully converted a big company to long-term client, laid the foundation for future projects, and paved the way for addtional clients.
-
Data Science Researcher
Institute of Information Science, Academia Sinica- Published an NLP paper in HCII 2019 related to supply chain fiannce.
-
MS Economics
Universität Mannheim- DAAD Graduation Scholarship
-
Double Degree Mathematics & Economics
National Taiwan University- Presidential Awards x2
- Chien Shih-Liang Scholarship
Highlights
Time Series
I possess expertise in various models and advanced techniques for time series forecasting (mixed-frequency model, hierarchical tuning etc.) and have successfully delivered two time series projects to my employers. I helped Forecasty.AI converted a big firm as its first long-term customer, and helped Lufthansa Technik gain much business attention on the project for decision making.
Natural Language Processing
I published a paper in 2019 to classify the commercial relationship of any given pair of companies in supply chains. This success is based on my solid knowledge of name entity recognition, knowledge graph, and domain knowledge. I now practice combining time series and NLP together to deliver better services.
MLOps
Not just the model performance, but also the model stability and monitoring plays a crucial role in earning clients' trust. Having worked in a DevOps team for machine learning service, I experienced developing end-to-end ML pipelines with continuous integration and delivery, including model monitoring and retraining for best practice.
Industrial Organization
I finished MS Economics competiton and regulation track, which builds firm theory of various forms of price and product competitions based on game theory. This secret weapon helps me not only on my daily work for data engineering, but also on business occassions when firm needs to take decision in response to prevailing market conditions.
Causal Inference
People become more prudent and prefer decoding model behaviours when it comes to decision-making relevant projects. Price and demand forecasting belongs to this category. With graduate training, I am able to implement causal inference and impulse response, helping companies interpret models and make most accurate decisions.
Advanced Skills
Reading paper, implementing innovative ideas, and collaborating with the DevOps team are part of my daily job. Leveraging my BS Math background, I develop efficient and clean code in short time, and have impressed my employers with my quick-learning ability and open-minded attitude.
LATEST WORKS
Master Thesis
I designed and experimented 10+ models for hierarchical flight hours forecasting. It outperforms company's existing model, gains lots of buiness attention, and lays the foundation for similar projects.The thesis is locked and protected by Lufthansa Technik.
Commodity Price Forecasting
By successfully delivering a model with less than 5% test error (orange line) and passing a 3-month live-testing period in the market, we fulfilled our customer's request for a chemical commodity's weekly price forecasts across three regions: US, EU, and Asia. This accomplishment not only helped Forecasty.AI secure its 1st long-term client, but also paved the way for additional clients. The project is locked and protected by Chemovator GmbH, the incubator of BASF.NLP Paper in Supply Chain Finance
The paper aims at classifying the commercial relationship between any given pair of companies in industrial supply chains: upstream, downstream, parallel collabaration, and no-relation. We proposed a novel word embedding scheme by building a multi-relational graph, outperforming Wor2Vec and GloVe. Checkout this PaperRefinitiv Crawler
I delivered the crawler to my client and efficiently performed robust crawling of thousonds of company data from Refinitiv. Notably, I received highly positive experience following the delivery, a testament to the quality of my work and dedication to customer satisfaction. Youtube DemoTestimonials
We all need people who will give us feedback. That's how we improve. — Bill Gates
Wen-Kai has quickly demonstrated deep knowledge in time series modeling and software engineering and therefore has stood out taking on additional responsibilities as a technical lead of our AutoML modeling pipeline and the beta release of our online platform.
We appreciate Mr. Chung for his dedicated attitude and ample contribution to our products. We love to work with him and strongly recommend him as part of your vision in the future.
Dr. Rahul K M
CEO, Forecasty.AIWen-Kai is a talented young data scientist with exceptional statistical and AI knowledge, capable of developing high-performing models to support business processes. His creativity, tenacity, and hard-working are valuable to solve complex problems.
Beyond their technical prowess, Wen-Kai is an invaluable team player. He is also mature enough even to work independently, and report results to the leadership.
Erika Fonseca
Delivery Solutions Architect, DatabricksIt has been a pleasure to work together with Wen-Kai. Apart from his impressive technical ability, he showed lots of enthusiasm and interest for any topic assigned to him and contributed greatly to the team with his willingness to help others and a collaborative mindset.
David Guo
Data Scientist, Lufthansa TechnikWen-Kai is a highly competent and diligent programmer and a good friend. The crawler he coded for me fulfilled all its goals and well exceeded my expectations. Thanks to him my research project could finally move forward. I would definitely love to work with him again in the future!
Minrui Gong
Doctoral Student, Universität MannheimOther Projects
Kaggle: Water Level Prediction
The competition aims at forecasting the amount of water for each waterbody in Italy. We apply VAR/VECM/LSTM models and give causal inference to find important features via Orthognalised Impulse Response (OIR) and Forecast Error Variance Decomposition (FEVD) technique. Kaggle (or, GitHub)Solutions to C++ How to Program
I finished exercises of most important chapters of the book C++ How to Program 4/e, Deitel et al. This is why I write clean and efficient code with strong OOP style in all programming languages. GitHubSolutions to Linear Algebra
I finished 95% exercises of the textbook Linear Algebra 4e, Firedberg et al., the standard book for mathematics department. It benefits me for all future careers and projects. GitHubCertificates
IBM DevOps and Software Engineering
The track contains 14 courses, professionalizing in Cloud Engineering and CI/CD, with a total of 160 learning hours. See credential for detailsDeeplearning.AI GenAI with LLM
With a total of 16 hours and 3 assignments, the course provides deep understanding of different strategies of imprvoing LLM output, e.g., 0,1,N-shot prompt engineering, full vs PEFT prompt tuning, and Reinforcement Learning with Human Feedback via PPO policy. See credential for detailsDeeplearning.AI Deep Learning Track
The track contains 5 courses, 20 projects, in total of 120 learning hours. See credential for detailsDeeplearning.AI NLP Track
The track contains 4 courses, 16 projects, and a total of 120 learning hours. See credential for detailsUdemy Aerospace Engineering
The course provides basic and advanced understanding of Aircraft design with a total of 4.5 hours. See credential for detailsDataCamp NLP Track
The track contains 6 courses, and a total of 25 learning hours. See credential for detailsDataCamp Three Career Tracks
It contains data scientist, ML scientist, and data engineer career tracks, comprising 65 courses, 6 projects, and a total of 254 learning hours. See credential for detailsGet in touch
Interested in my profile? I am always happy to receive your message and open to discussing new projects, creative ideas or opportunities to be part of your vision!