Sandipan Dey

Working in the insurance domain for a product named BANCs, usecases such as Health-insurance assistant – a voice bot with amazon alexa; A chatbot with Amazon Lex using Facebook Messenger channel; Using Amazon Textract / Azure Form-Recognizer for OCR; Next best action recommendation, policy / fund-switch recommendation, fraud detection, automated processing, cross-selling / up-selling of products.

Worked in events domain for a product (with wearable beacons and near-real-time analytics), data science usecases such as User-matching / match-making with demographic / behaviour features and using OKCupid / ELO scores; Personalized recommendation with user-user / item-item CF, CBF and Neural CF algorithms; User segmentation; Most frequent booths / sessions visited together with apriori; Predicting availability of booth representatives with logistic regression and ensemble models. Also worked on pre-processing noisy RSSI data, using smoothing algorithms, trilateration algorithms, and kalman filters.

Worked on claim prediction for medical insurance, developing predictive models using Time series ARIMA / SARIMA models and ML based models like SVM, RandomForest with lags along with user segmentation with Kmeans clustering. Also worked on aspect-oriented sentiment analysis product, feature extracted with shallow NLP parsers such as SDP.

Worked on a POC for a bank, building predictive models for deposit, EOD balance, loan amount with decision trees and random forests. Customer Segmentation with hierarchical clustering. Also worked on sentiment analysis on twitter feeds, working with twitter / facebook graph API. Working on Polarity / Sentiment Analysis / Emotion Classification / Topic Segmentation. Working on email mining.

Worked as a consultant/information architect. Helped clients use advanced analytics models. Worked on M2M Advanced Analytics use-cases, building predictive models for device failures (prognostics/predictive maintenance). Used various ML techniques (Sequence Mining, Apriori, Random Forest, SVM, KMeans, PCA, etc.) and statistical models (ARIMA, SARIMA, Kalman, Survival analysis).

Helped clients use hadoop echosystem / BigData efficiently and develop data science applications on top of hadoop. Used Amazon EC2/S3/EMR to parallelize NMF computation system and hive queries for pattern matching on BigData.

Implemented a demo web-based multiple-choice pharmacology-domain question answering system as POC. Worked on study of distance measures and ranking methods for finding similarity of concepts. Enhanced a C# winform-based GUI tool for knowledge-encoding.

Worked as a developer in the dev team of a V1 product in agile software development model (scrum). Implemented new features for different modules (client/backend/UI). Worked on FxCop.

Worked as Team Lead of Batches & Reports module and developer. Developed a generic C Utility for CSV Report generation, worked on performance optimization.

Worked as application developer. Enhanced/implemented features (e.g., subgroup feature, crosstab feature). Implemented a license-key generator.

Worked in VCS Front-end / Middle-end. Enhancement, coding, writing test-cases, regression testing. Developed small POC tools.

Developed and maintained CYTOSOFT 2.5. Ported the multithreaded application from Windows to a platform-independent version using Qt/STL/C Runtime.