CakeResume 找人才
最輕量、快速的招募方案，數百家企業的選擇
搜尋履歷，主動聯繫求職者，提升招募效率。
- 瀏覽所有搜尋結果
- 每日可無限次數開啟陌生對話
- 搜尋僅開放付費企業檢視的履歷
- 檢視使用者信箱 & 電話
職場能力評價定義
Kao Chiang
NTU ME
#ML #NLP #Recommender
Skills
Software
> ML: Tensorflow, Keras, Xgboost,
LightGBM, Pytorch
> NLP: Rasa, spaCy, Gensim, GluonNLP, GloVe
> Web: Flask, Django, Vue, SQL
> APP: Swift, Android Studio
> Others: Docker, docker-compose, Spark
Combination
Communication
Nowadays, it is quite easy to use many powerful tools, but mostly each of them are used in different platform or languange, and each of tools often include many knowledge which need to be considered various aspects. In some experience of system design, I
Projects
Chatbot
An end-to-end chatbot platform which includes two part of main services. One is a friendly interface to edit corpus and dictionary for training machine-learning model; Another platform is a chatbot services include chat room for testing , logging for remarking incorrect response, and so on. Most of NLP models are applied in English or western language, but our clients are Chinese. So, I need to re-write many program flow to be suitable and well on Chinese .Face Recognition
In a corporation with a security company, they want to include some AI in their security system in an exhibition. There are two main of conditions. One is to apply in department store to recognize and record the flow of people with gender and age instantly. The other one is an access control by recognize face of people.
To their demand, we make a device embedding two machine learning model to detect face and recognize age and gender.
Recommender System
It is lucky to participate the design of recommender system of a top e-commerce platform, and that is the first time I took "big data". Because of the amount of data, the data pipeline need to be very careful in the parallel computing. We use Apache Spark framework and Kubernetes to deploy our models.
The recommender system is mainly combined by two models. One model is user-based model and the other is content-based model. We use fully-connected layer to combine.
Kao Chiang
NTU ME
#ML #NLP #Recommender
Skills
Software
> ML: Tensorflow, Keras, Xgboost,
LightGBM, Pytorch
> NLP: Rasa, spaCy, Gensim, GluonNLP, GloVe
> Web: Flask, Django, Vue, SQL
> APP: Swift, Android Studio
> Others: Docker, docker-compose, Spark
Combination
Communication
Nowadays, it is quite easy to use many powerful tools, but mostly each of them are used in different platform or languange, and each of tools often include many knowledge which need to be considered various aspects. In some experience of system design, I
Projects
Chatbot
An end-to-end chatbot platform which includes two part of main services. One is a friendly interface to edit corpus and dictionary for training machine-learning model; Another platform is a chatbot services include chat room for testing , logging for remarking incorrect response, and so on. Most of NLP models are applied in English or western language, but our clients are Chinese. So, I need to re-write many program flow to be suitable and well on Chinese .Face Recognition
In a corporation with a security company, they want to include some AI in their security system in an exhibition. There are two main of conditions. One is to apply in department store to recognize and record the flow of people with gender and age instantly. The other one is an access control by recognize face of people.
To their demand, we make a device embedding two machine learning model to detect face and recognize age and gender.
Recommender System
It is lucky to participate the design of recommender system of a top e-commerce platform, and that is the first time I took "big data". Because of the amount of data, the data pipeline need to be very careful in the parallel computing. We use Apache Spark framework and Kubernetes to deploy our models.
The recommender system is mainly combined by two models. One model is user-based model and the other is content-based model. We use fully-connected layer to combine.