Prediction of Diabetes from Lumiata Claims dataset

• Streamlined ETL operations on a semi-structured Claims dataset and analyzed the important variables i.e. LOINC, ICD 10 and, CPT codes. Balanced the data using stratified sampling and predicted the probability of diabetes for patients using SVC model achieving a recall of 89%.

Identification of Leader words from any local meaning

• Optimized Python script to transform unstructured data from Merriam Webster dictionary to structured format a strong rule-based association. Modeled a Universal Leader words identifier using Bi-directional LSTM which takes Glove Embedding and POS tags attaining a recall of 96%.

Sentiment Analysis of Product Reviews

• Implemented a rule-based classification and a Deep Learning LSTM model for sentiment analysis of 4 different languages i.e. English, Spanish, German, and French. The rule-based model outperformed the LSTM model achieving an accuracy of 92%.

Twitter Topic Modeling Application

• Developed an application that applies topic modeling on the tweets of a subject using unsupervised LDA, and semi-supervised CorEx. Obtained 10 different latent topics and their sentiment over 14 weeks. Deployed the model in production using Azure ML deployment

• Performed statistical analysis of FI and SD data and suggested tangible solutions to stakeholders by visualizing the KPI metrics in Tableau.

• Streamlined workflows and built applications in SAP business modules deploying the applications using the IBM TWS tool.

• Optimized complex SQL Queries for faster data processing which improved the speed and efficiency of applications by 28%.

• Key achievement: Received “Pat on the Back Award” for spearheading the whole team and managing adhoc requests during challenging time.