江于萱
I majored in Probability theory, and focused on Markov Chain related problems. After graduation, I worked at ASUS on Zenfone camera algorithm development for 2 years. Then focus on self-learning platform "Coursera". I completed the series of class "Advanced Machine Learning". Now, I am a data scientist in KKday. The main projects are listed below :
- Consultant for Marketing Department : already analysis over 20 topics for optimizing allocation of advertising expense.
- Personalized city recommend on KKday homepage for increasing click rate over 1%.
- Decreasing fraudulent for risk controlling department.
Machine Learning Engineer / Data Analysis Engineer / Data Scientist
Taipei，TW
[email protected]
Skills and Certifications
Language
- Python
- C++
Mathematic
- Probability
- Stochastic Process (Markov Chain)
- Probability Model
Work Experience
Data scientist at KKday : 1.5 years (2018/09~ )
Consultant for Marketing Department Objective: Optimize allocation of advertising expense (CID/EDM)
- Coupon
- Repurchase rate
- Customer behavior by locale
- Purchase platform (App/Mobile Web)
Consultant for Risk Controlling Department
- Fraudulent (3D verification)
- Study the solution of other E-commerce
Personal profile features on KKday.com
- Design city recommendation system (based on cookies) and launch this feature on KKday homepage since 2019/11. (Click rate increase over 1%)
- Design personal database for future project
Self study : 0.5 years (2018/03 ~ 2018/08)
Study Advanced Machine Learning Specialization series classes ( 7 classes )
Software engineer at ASUS : 2.5 years (2015/10 ~ 2018/03)
Camera developer for Zenfone
- Design white balance algorithm (Based on RGB sensor) and launch this feature on Zenfone 3/4 series successfully.
- Design camera automated manufacturing tool and has been deployed to factory (located at Suzhou, China) for producing Zenfone 3/4 and Zenbo successfully.
- Analysis the defeat of vendor (ex: Camera module, Samsung) and issues have been fixed successfully on Zenfone 4 Pro.
Student : ~ 2015/06
Master&Bachelor : National Chiao Tung University , Applied Mathematics
KKday
Personalized city recommendation on KKday homepage.
Based on language , customers purchase history and recently action, we generate cities interesting to customers.
This model increased click rate over 1%.
Data Analytic - Purchase Platform
The target is comparing the value of customers who buy products on different platform.
I analyze from three aspects :
- Average order price on different platform
- Average value of customers with first order on different platform
- The preference of customers who have ordered on multiple platforms
Kaggle
This is a time model. For each time, it has only 5 features (shop_id/ item_id/ Category_name/ shop_name/ Category_id), So I need generate new features. I add time delay data and embedding the shop name and item category as new features. I also train model to get new features. In the end, I use Linear model to reduce features and build 3 models to get the final answer.
Project 1 : Image Captioning
Given a picture, It will generate a short description for this picture.
This model use a pre-trained InceptionV3 model for CNN encoder and extract its last hidden layer as an embedding.
Use about 82K training data and 40K validation date and each picture has 5 captions.
Project 2 : AI robot in Telegram
If you ask a programing problem, chat robot will return a closest StackOverflow link for you.
In this model, we will classify the training data to language type and use facebookresearch/StarSpace for embedding questions. For each input, we will classify it and embedding it to vector then found related link.
Project 3 : OpenAI CartPole-v0
Use Monte Coral tree search. That is, we choose road by root scores, if meet the tree leaf, do propagation (add score to root).
We can build a tree by playing games, then tree will tell us how to playing game.
Project 4 : Playing Game
Use actor-critic training.
Input 4-frame image, train a DNN to catch the image information and predict the probability for 12 reactions. Then we according the DNN result to make a reaction for this game.
Master Thesis - A Random time for Simulating Markov Chains
In my master thesis, it provide a simulated method, which can avoid lots of computations, to make the Markov chain approximate its stationary distribution and also give a theorem to prove. At first part, we gave a theorem to prove the convergence of new random variable. For second part, we gave two special cases of simulation and found the random variable will not converge if the chain does not satisfy the condition of theorem. In the end, we provided a way to improve the chain.
