Groww, India

Major responsibilities1. Predicting Customer churn and identify factor driving customer’s churn:a. Trained different tree based models like Lightgbm, XGboost to predicts user churn probability. b. Used Shap library to identify the important factors driving the churn of customers and explain the model outputs.c. Used Shap dependence plot (sensitivity plots) to identify the percentage change of churn with respect to feature values, so product and business teams can perform… Major responsibilities1. Predicting Customer churn and identify factor driving customer’s churn:a. Trained different tree based models like Lightgbm, XGboost to predicts user churn probability. b. Used Shap library to identify the important factors driving the churn of customers and explain the model outputs.c. Used Shap dependence plot (sensitivity plots) to identify the percentage change of churn with respect to feature values, so product and business teams can perform campaigns for increasing customer’s retention. 2. Keywords generation for FAQa. Trained different seq to seq transformer models like BART, T5 etc for generating keywords for FAQ’s.3. Topic modelling for segmenting NPS comments:a. Used BertTopic library to create topics using Bert(Transformer) embeddings for comments written in nps survey in an Unsupervised approach. 4. NPS Survey comments classificationa. Trained different text classifiers like (BiLSTM, Bert, Roberta) for classifying the comments written in NPS survey comments.5. Customer Tickets classification.a. Trained different nlp models from BOW to transformer models to train a multiclass classifier to classify tickets raised by the customers.6. Semantic search model for help and support section:a. Fine tuned a paraphrase model to fetch relevant FAQ’s in help and support search bar. b. Converted the model to onnx format and achieved an inference time of 3 ms for encoding a query for better customer experience.