Seeing the North Sea @ Castricum aan Zee in October, 2015
Hello! I am Raymond Kwok, and I am from Hong Kong.
Being a business-oriented Data Scientist with software development background, I am experienced in customer analytics using causality, A/B testing, machine learning methods, programming and so on. With them, I fought my most unforgettable battle in 2018 in which I successfully boosted Lalamove's revenue in China by 10% month-on-month.
This webpage will tell you more about me, especially my career, along with some of my user-oriented projects. I will be delighted to have further conversation with you.
Feel free to contact me.
From R&D
FROM R&D
I started off my R&D career at the early stage of a frontier neutrino experiment inside a nuclear power plant, working with people from the US, China, Russia, Japan, and so on.
It was the right job at the right time because it broadened my horizons and gave me unlimited room to strive for the best, polishing my skills in data analysis, software and hardware implementation.
One day, my boss tasked me a sub-system, co-leading the design, construction, and maintenance of a gas supply system with a $60k budget to keep the neutrino detectors from the air and water. I presented my design and progress in many meetings, contacted 30+ parts suppliers, learnt how to bend a copper pipe, and how to make quality mechanical drawings, and so on.
I could not forget the moment, after days (underground) and nights (in hotel room) of endeavor, when the chief engineer looked at my work, the only word he said after a long silence - "Amazing".
Video made by me in 2015 to introduce the principle and facilities of the experiment for public engagement.
(Credits to Jonathan Wong for the English voice-over)
LINKS
- Paper: Observation of electron-antineutrino disappearance at Daya Bay |
- Announcing the First Results from Daya Bay: Discovery of a New Kind of Neutrino Transformation |
- 2016 Breakthrough Prize in Fundamental Physics |
- Scientists Say Farewell to Daya Bay Site, Proceed with Final Data Analysis |
- Video made by me to introduce the principle and facilities of the experiment |
To Data Science
TO DATA SCIENCE
After another 2 R&D roles in industry, I decided to pursue a different way to innovate - Data Science. While limitation in Hong Kong was a factor, my background, the rising of the field, accessibility to infrastructure, and growing of the community were all my driving forces.
In 2018, I modeled Lalamove's customer data with Bayesian statistics, and built an A/B test system to run a daily campaign so that I could react to the market response data-driven-wise. We made a 10% MoM increase of revenue. Taking advantage of the company's strong foundation, my approach allowed us to fine-tune our strategy for a better and better return-on-investment.
To equip myself with the latest technology, I completed a 5-course specialization in Deep Learning (organized by deeplearning.ai of which now I am an alpha test consultant), and finished a trip planning recommender for the capstone project of the IBM Data Science Professional Certificate in 2019. I had three other moves in the same year.
First, I started to write data science blog, and shared my articles with Analytica Vidhya, receiving ~38,000 views and a 50% read-rate so far. Secondly, I began my Master of Science degree study (specialized in Data Science), offered by the University of California, Riverside. And lastly, I moved to Montenegro, a beautiful European country, for my first official Data Scientist position.
LINKS
Projects
Heuristic learning with triplet network (2021)
for transfer-airport finding in interlining
Achieved over 90.3% recall rate in learning transfer airports options based on > 1 million rule-based one-stop journeys. The method could be applied to learn users' patterns for making personalized interlining journey suggestions.
Product search by photos (2021)
5th place in an eBay competition and presented in the FGVC8 workshop
Trained Neural Networks with > 1 million photos to use a photo of a product to search for the products' other photos that were taken from different angles or in different places. The technique could be useful for product price-matching or finding relevant user posts.
Feature selection with Causality (2020)
for a Portuguese bank telemarketing dataset to predict customer conversion
Simple yet robust model could be built by selecting features close to the predicting variable in causal graph.
- Project poster |
- Blog |
Web crawler and Search engine (2021)
Crawled webpages, extracted and indexed their main content for the use of a search engine. It demonstrated the idea of incorporating user's preference into search results.
- Git repo for crawler with Readme for description |
- Git repo for search engine with Readme for description |