• As a part of the Advanced Analytics – Fuel team, involved in creation of data pipelines to extract, transform and load fuel data in Palantir Foundry using Ontology Layer, Code Workbooks/Workspaces, and Contour Boards• Developed and deployed Time-Series(ARIMA, SARIMAX), OLS, Random Forest, Sequential NN, and Gen AI models including BERT tuned to predict fuel burn, and uplift 2 weeks in advance for UA/UAX operated flights throughout the world with >95% accuracy and WMAPE <1%• Continuous monitoring of existing legacy fuel burn model deployed in Ontology layer to track model performance using business defined metrics by week, month, and quarter.• Analysis of United’s fuel uplift records vs. vendor receipts to identify discrepancies in fuel uplift using Contour boards that helped stakeholders understand where fuel was being mishandled • Created datasets using Code workbooks and Contour boards from different source tables in Palantir Foundry to use as an input for predictive models to predict fuel uplift.• Utilize Python, Pandas, Numpy, and Pyspark to clean, preprocess, and analyze large datasets, ensuring data quality and integrity.• Apply machine learning algorithms from Scikit-Learn and Statsmodels to develop predictive models for various business problems, including classification and regression tasks.

• As a part of the marketplace decision engine and auction team, developed and deployed predictive models to grade and route premium Medicare eligible leads(A+ leads with high CRs) to our partners resulting in a monthly revenue addition of >$100,000• Developed and deployed scalable and high performance predictive ML algorithms (including Random Forests, LGBM and RNNs to predict Policy cancellations with >95% Recall and Churns with 95% precision) and survival models (including Cox Hazard, Kaplan Meier, and NGBoost) to predict Customer value to identify the most profitable demographic to target sales of Medicare Advantage policies during by plan and carrier and leading to an increase in revenue of >$500,000• Generated Synthetic Data from historical CMS data using GANs to predict the LTV (lifetime value) of a customer with accuracy >90% and low RMSE• Developed and deployed predictive ML models to identify the target audience (based on Customer behavior data) for our Facebook Ad Campaign and used it to automate Bid Multipliers to reduce ad spend by 3x and increase ROAS by 5x• Pioneered ad-hoc projects for the marketplace and finance teams to identify and visualize lead characteristics (by region, device type, age, gender etc.), to sell 5-STAR Medicare Advantage Plans to throughout the year (searchable by plan, carrier, zip code and deployed on Streamlit) that reduced average CAC by 80% and customer drop-off rates by 3x• Setup experimental designs on AB Smartly to test and validate the performance of the above-mentioned AI/ML models daily

• Analyzed Home Healthcare Patients’ data throughout the USA to develop and deploy machine learning models including Logistic Regression, Random Forest, XGBoost and Neural Networks to predict Hospital visits/Emergency care with a precision of >90%• Developed NLP models based on causes identified for hospitalizations to suggest improvement measures which reduced hospital visits by 2x and increased agency revenue by more than $250,000

Developed Deep Learning NLP models to detect malware in a wide range of text files including batch files, JSON files, text files, PowerShell scripts

Teaching assistant for the Regression Analysis course at Rutgers University, New Brunswick

Hotspot detection, analysis, and modeling of grade crossings in ArcGIS where there are a lot of accidents by number or rate, by maintenance that has/has not been done in locations/regions,

Analyzed Segment and Supply Chain Performance based on firm-level sustainability assessments for the year 2019 to propose future actionable objectives of risk reduction analysis at products/services level and cost reduction analysis due to sustainable procurement practices. Assisted with data analytics for public disclosures on sustainability/corporate social responsibility issuesand the supply chain, specifically related to climate change, human rights, and diversity and inclusion.Analyzed key impacts of current Nielsen supply chain strategic suppliers.

• Analytical research project focused on brand defects, data sets, business insights, and qualitative analysis all impacting decision making• Standardized the values from current OPERA system and OLQ process through Visualizations in Tableau and designed the Global Defects Library with images and search capabilities

Modeled insurance risks for both Pre- Quote and Post-Bind accounts (CP, EER and PR) for the American regions (Southern, Midwest and Mid-Atlantic) and Canada.Analyzed various Beta distribution curves/ EP curves which determine the Technical CAT Premium (TCP) using tools such as Microsoft Excel, Microsoft Access, VBA, R scripts and in-house tools such as Risklink, Riskbrowser and GES.Successfully worked with the Underwriters in USA and assisted them booking important accounts with insured values more than $100B and premiums greater than $3M USD.Prepared Risk insights and Exposure Summaries using Pivot tables and charts in Excel and R scripts for a lot of new accounts, each of which were presented to the client by the Underwriters in USA which helped us book these businesses and increase our overall portfolio by $15M, providing us an edge over our competitors.Worked on end-to-end processing of accounts which reduced the turnaround times by almost 60% without compromising the data quality for rush accounts which needed expedited processing.Used GIS for data enhancement and risk engineering to inform clients about the regional risk profile.Managed the entire Post Bind team responsible for gathering and verifying data used to create the portfolio of $6.5B in premiums.Solely responsible for updating accounts into the Access tracker daily and delegated these accounts to team members based on their skill set of account processing quality and speed.