Dec 2021 – Sep 2024 · 2 yrs 9 mos · Singapore · On-site

Jan 2019 – Dec 2021 · 2 yrs 11 mos · Singapore

Jan 2018 – Dec 2018 · 11 mos

Jul 2016 – Dec 2017 · 1 yr 5 mos

• Played a major role in end to end implementation of Deep Learning, Support Vector Machine and Multinomial Naive Bayes models, on a distributed framework, for performing information extraction from un-structured legal documents. This model will help agents improve efficiency by up to 50%.
• Applied transfer learning, using pre-trained VGG16 and RESNET50 deep learning models, trained on IMAGENET data to identify lung cancer from a repository of DICOM images.
• Worked as a part of a two member team to classify news articles as relevant or irrelevant based on pre-existing tags provided by the operations team. Also performed un-supervised clustering using Latent Drichlet Allocation to gauge the different themes present in the article. This model was integrated as a web tool to further assist agents with the relevant news identification.

Jan 2015 – Jun 2016 · 1 yr 5 mos

• Applied an ensemble of GBM and Random Forest algorithms over marketing campaign data (anonymized) of a major US firm to generate scores which determine the probability of campaign response.
• Using the principles of Network analytics developed a method to identify fraudulent physicians from a network of physicians operating in the United States.
• Worked as a part of three member team to develop a module in pyspark to detect online misuse.
• Used Amazon Web Services (S3, EC2, Elastic Search, Lambda, Kibana) to develop an end to end solution to continuously monitor tweets containing pre-specified keywords. Final summary and plots were rendered using Kibana.
• Developed an algorithm to analyze Workers Compensation and General Liability claim notes of a major US client. Based on the theme of the notes, provided recommendation to the respective party involved.
• Implemented a recommender engine using market basket analysis on the bank transaction data to identify customers who have propensity to buy Life Insurance products. As a precursor to this project, also performed clustering of customers into eight prominent segments using K-Means clustering.

Aug 2013 – Dec 2014 · 1 yr 4 mos

• Developed a logistic regression model in SAS to detect early life insurance claims. This model helped underwriters prioritize the claim underwriting process, at the same time, preventing false claims worth millions of dollars.
• Performed a study, comparing advantages and disadvantages of using machine learning techniques such as Random Forest and GMB over traditional statistical modeling techniques, over the life insurance claims data.

Jan 2010 – Jun 2011 · 1 yr 5 mos

• As a part of the six months computer science training program developed an thorough understanding of OOPs concepts along with web technologies.
• Was responsible for UAT to production migrations in the Siebel environment.