Shrish Chandra Pandey

Initiated automated extraction of qualitative insights from YT and Twitter comments on a released content. Finetuned Roberta based sentiment classification model and employed topic modelling to assist consumer insights team assess content reception. Currently working with open-source LLMs for summarizing the reviews for easy consumption. Developed and deployed a CatBoost powered Winback Propensity Model for churned SVOD customers enabling model explain-ability using SHAP values. Dealt with highly imbalance data and got a precision of 0.72 and a recall of 0.83 on the minority(winback) class helping the marketing team curate outreach strategies.

Developed a Lead Segmentation model for business funnel optimization. Tried various clustering approaches and converged with KMeans to come up with 6 lead cluster profiles to figure out motivations to join a learning program. Observed statistically significant growth of 0.4% at the Lead-to-Application stage of the funnel for 2 offerings. Developed various propensity models for marketing and operational usecases. Worked as a content contributor for AIML and DSBA programs, learning and working on wide range of case studies with exciting use cases. Also, collaborated with US universities(UT Austin, MIT IDSS) for curriculum design for new product development.

Developed an NLP powered Customer Experience Tableau Dashboard to churn out insights from complaint call data generated for a fortune 100 financial services client. This included- An information retrieval system based on QnA transformer models to extract the themes in complaint call transcripts. Pipeline included data specific embedding, Question Answering Model, Clustering (DBScan), POS based preprocessing to uncover hidden themes. A Zero-Shot Classification model mapping back identified themes to a call. Generated a mapping function to map embeddings in a large vector space(BERT) to a small vector space(Word2Vec) to reap benefits of both the embedding techniques. Worked on Anomaly Detection, Time Series and propensity modelling use-cases for FMCG and Finance domain.

Got trained on .NET layered architecture along with SQl, Python, C# and Cloud computing.