-- Building recommender systems in the healthcare space

-- Launched the first provider recommender system across multiple channels which will help members find the best doctor in 26 clinical specialties based on their personal preferences. New recommendation system is expected to show 144% LIFT in user satisfaction and reduce medical cost by 15%.-- Collaborated with DE and UI team to design and set up a new provider navigation tool which is expected to reduce time to action by 60% for end users.-- Realized the need for improvement and acted on it by leading a new data pipeline design for the data science team by introducing a modular code structure, which solved scalability issues in the previous codebase. Achieved 85% reduction in query time on-prem.

-- Implemented a new optimization algorithm as a part of Aetna’s next best action campaign which will generate a specialist referral list for PCPs based on member utilization and create a recommended list of specialists for elective surgeries for members. The campaign is predicted to generate $8M in annual savings and increase member satisfaction by 20%-- Created production data pipeline using PySpark to ingest 8TB claim data of 19M members which are used to train a tree-based model using XgBoost to predict members’ willingness to travel based on utilization pattern which improved prediction accuracy by 30%-- Designed a cluster-based algorithm using Unsupervised ML models Hierarchical Clustering, DBSCAN to identify member level geo-preference for a doctors appointment-- Developed recommendation algorithm to find out preferred providers for 51 HealthHUBs across 8 specialties based on demographics and several other determinants which will generate $4M annual savings-- Wrote a Python package which will test our KPI on a quarterly basis

-- Discovered a new technique of digital holography in collaboration with 2 other Ph.D. students, which is expected to increase NASA’s LiDAR laser system power efficiency up to 33%-- Created a machine learning algorithm by implementing deep convolutional neural network using Keras in Python to analyze properties of colloidal solutions-- Wrote algorithms in Python to analyze over 500K images of flow-channel to find out the particle concentration in a solution using Shannon entropy-- Implemented Levenberg-Marquardt algorithm for particle tracking and streamlined using TensorFlow-- Used Fast Fourier Transform to create exotic modes of light, which has optical communication applications-- Selected from >300 students to represent NYU’s Soft Matter Research program during the NSF Site Visit to give a poster presentation, which led to a $10M research grant renewal-- Spearheaded the experimental setup and build-out of a new research laboratory.-- Trained 3 undergrads to perform an exhaustive test on a research project.

-- Instructed weekly recitations for more than 150 science majors and almost 100 non-science major undergrads, ensuring proper understanding of all topics, and as a result their test scores improved from an average of B+ to A--- Outlined and offered lessons before examination periods, including help sessions for individuals and for groups.

Insight is a seven-week postdoctoral training fellowship that bridges the gap between academia and a career in data science. The program enables scientists to learn the industry-specific skills needed to work in the growing field of big data at leading companies.-- Developed “Smart Hotels” (http://smarthotels.city) a web app to help hotel managers prioritize where to invest their resources to improve their hotel by analyzing customer reviews leading to lower operation cost-- Trained a Mallet LDA model on customer reviews scraped from Booking.com which can determine the topics discussed in each line of a customer review-- Designed a VADER sentiment score analyzer in order to compare the quality of hotels across each discussed topic among hotels with similar star ratings

-- Practice case-based interviews together with advanced degree students with similar interests

-- Created a tree-based model using Random Forest, and CatBoost which increases West Nile Virus prediction score by 26% for Chicago Public Health department using previous WNV outbreak and weather data-- Ranked within top 10% in McKinsey Analytics Online Hackathon, co organised by Analytics Vidhya for finding out the probability of a client having stroke based on their lifestyle, demographic and several health details.-- Ranked within top 25% in G-Research 'Financial Forecasting Challenge' by creating an ensemble model using Gradient Boosting Regressor, XGBoost, and LSTM.-- Implemented Convolution Neural Network to build an app for dog breed classification, using Keras backend to TensorFlow.

-- Developed the penetration strategy for a foreign beverage company by testing the product adherence in US-- Explored viable distribution channels to ensure the company’s target of annual EBIT increase by $12M-- Presented the summarized analysis of the business problem, developed by our team, to the stakeholders in the bi-weekly meetings

Solved a real world case with present consultants at New York office

-- Spearheaded a team of 5 graduate students to provide medium risk strategy for GameStop to increase annual revenue by $4.5b in next 5 years, which was selected as one of the 20 semi-finalists