Data Scientist

SCUBA for automated data-labelling:
* Created a tool which used open-source API to facilitate automated labelling called Snorkel. It allowed users to write rules which could be as simple as pattern matching and as sophisticated as running model. Once the rules were defined, we trained a weak supervision deterministic label model. The automated label generated from these rules could then be used to train a full-scale binary classification model. In additions to automated data labelling, we also added exciting features to augment data

Churn Analysis:
* Created churn analysis model for the HR team and provided them with the strategic Retention Plan and categorized the employee to various Risk Category. Association rule mining was also implemented to provide combinations of top manager, skill level and sub-band level etc under which employee was more

likely to churn.