Abhishek Sadhukhan

Developed and deployed RoBERTa model into GCP using Vertex AI and GKE. Built MLOps CI/CD pipeline to remove the manual invocation of various services needed for E2E model deployment and automate the workflow using Cloud Build. Achieved F1 score of 78.8 for RoBERTa.

General experience covering roles including Data Scientist, Data Analyst, and Test Engineer. Key achievements include performing Exploratory Data Analysis, Feature Scaling, and Data Visualization using Seaborn; building Regression models like Linear Regression, XGBoost; and determining key factors contributing to customer churn.

Objective: To build a predictive model that feeds on historical sensory data to find the expected failure rate of the encoder component of X-ray scanners and eventually determine the survival hours before teardown.

Objective: To reduce the manual and repetitive effort of the long review time for legal contract documents and eradicate errors, to build an automated workflow to identify and annotate the important clauses in legal documents.

Objective: To identify the key aspect terms from vast amount of textual data and predict the sentiment based on the hotel reviews on the website of an US travel company to help them generate more revenue by focusing on the important entities. Used Word Embedding technique Tfidf and and VaderSentiment to calculate the compound score for visualization. Explored LDA and applied to the data to perform Aspect Extraction and visualized the topics using pyLDAvis. Built a BERT model to perform Aspect level sentiment analysis to classify the hotel reviews.