Sr. Data Scientist at KKLab, KKBOX Group
Taipei, Taiwan
四月 2020 - Present
Taipei, Taiwan
- adapt Aveva PRiSM to a tier 1 company in Taiwan for anomaly detection on their producing pipelines.
- engage in a venture building project related to NLP and develop its algorithms and models.
- win 1st place in IMBD 2020 which is a competition related to manufacturing.
- maintain and develop KKBOX opinion mining system and sell it to other companies (SAAS).
九月 2018 - 四月 2020
Taipei, Taiwan
- build public opinion mining system by novel NLP models.
* build novel NLP models such as BERT-BiLSTM-CRF, ConvNet and TNet by TensorFlow.
* maintain, design NLP pipelines for opinion mining system.
* build the frontend page which can monitor artist's social volumes by React.
* build restful api by Flask, gunicorn.
* deploy this microservice by docker stack (docker-compose)
- maintain app's event log specification and analyze user actions.
- win 3rd place in KKBOX inhouse hackathon. we build a social network app called KKMeet which can recommend you new friends by your music preferences.
2016 - 2018
Thesis: Lyrics- and Audio Features-based Hit Song Prediction
Sr. Data Scientist at KKLab, KKBOX Group
Taipei, Taiwan
四月 2020 - Present
Taipei, Taiwan
- adapt Aveva PRiSM to a tier 1 company in Taiwan for anomaly detection on their producing pipelines.
- engage in a venture building project related to NLP and develop its algorithms and models.
- win 1st place in IMBD 2020 which is a competition related to manufacturing.
- maintain and develop KKBOX opinion mining system and sell it to other companies (SAAS).
九月 2018 - 四月 2020
Taipei, Taiwan
- build public opinion mining system by novel NLP models.
* build novel NLP models such as BERT-BiLSTM-CRF, ConvNet and TNet by TensorFlow.
* maintain, design NLP pipelines for opinion mining system.
* build the frontend page which can monitor artist's social volumes by React.
* build restful api by Flask, gunicorn.
* deploy this microservice by docker stack (docker-compose)
- maintain app's event log specification and analyze user actions.
- win 3rd place in KKBOX inhouse hackathon. we build a social network app called KKMeet which can recommend you new friends by your music preferences.
2016 - 2018
Thesis: Lyrics- and Audio Features-based Hit Song Prediction