Experience
January 2025 - Present
I am currently working as a Data Scientist in the Vaccine Team at a leading pharmaceutical product company. In this role, I apply advanced data analytics, machine learning, and statistical modeling to drive impactful insights that enhance vaccine development and distribution strategies.
Projects:
1. PIECON (Patient Engine for Consumer): Every vaccine campaign has the same problem that millions of eligible patients, limited budget and no clear way to know who is actually ready to act. PIECON was built to solve exactly that.
-
Developed a PU model using XGBoost to predict vaccine propensity across a large consumer population, addressing severe class imbalance problem by treating unvaccinated records as unlabeled rather than confirmed negatives- enabling the model to learn meaningful signal from a highly skewed dataset.
-
Engineered features across three data pillars namely claims, demographics and life style features and evaluated model performance using recall and AUC, aligning evaluation metrics directly with campaign targeting efficiency.
-
Built interpretable consumer personas using Decision Tree layered on top of PU model probability score, translating raw predictions into named audience profiles with distinct outreach playbooks covering digital ad strategy and HCP engagement priority.
-
Validated personas through lift analysis, measuring vaccination concentration within each segment relative to the overall population rate and designed a closed loop feedback system where campaign outcome continuously retrain the model, improving targeting precision and lift with every cycle.
2. Uplift Modeling: Optimize multi-channel pharmaceutical outreach by identifying and prioritizing the exact consumers whose treatment decisions are genuinely driven by our engagement.
-
Differentiated "Persuadable" individuals from those who would take action independently ("Sure Things") or remain entirely unresponsive, preventing wasted communication budget.
-
Developed an uplift modeling framework leveraging Meta-Learner architectures (S-, T-, and X-Learners) to estimate Individual Treatment Effects and score personal responsiveness.
-
Built high-value audience personas for the persuadable cohort by analyzing their distinctive demographic attributes and lifestyle indicators, giving marketing teams clear behavioral profiles for targeted messaging.
-
Evaluated incremental conversion lift across each distinct persona segment, empowering commercial teams to deploy high-impact outreach strategies that maximized campaign efficiency
March 2022 - December 2024
I worked in digital analytics (pharma) domain as a data scientist at Genpact. I have experience in customer journey analytics, market mix modeling, omni channel analytics etc.
Projects:
1. Customer Journey Analytics via Pathways:
-
The project aims to identify marketing interactions effective at different stages of customer journey.
-
Defined success metrics for various stages of the customer journey (Awareness, Consideration, Engagement, Loyalty, Advocacy) through data-driven analysis and collaboration with brand teams.
-
Utilized community detection algorithms to identify key patterns and success metrics within customer journey stages, enabling more targeted marketing strategies.
-
Developed and implemented data preprocessing techniques to prepare extensive marketing data, prescription data and web traffic data for analysis in a customer journey modeling framework.
-
Identified and prioritized key marketing drivers using Elastic Net modeling, narrowing down from fifty potential predictors to the most impactful ones.
-
Utilized advanced clustering techniques (KMeans, GMM, etc.) to segment marketing drivers, enhancing the understanding of customer interactions and behaviors.
-
Developed regression models using XGBoost to identify the main marketing variables driving desired outcomes at each journey stage. Utilized SHAP (Shapley Additive exPlanations) values to pinpoint key marketing factors influencing success metrics.
-
Employed Structural Equation Modeling (SEM) to construct and analyze customer journey pathways, determining critical nodes and optimizing marketing efforts at each stage.
-
Calculated success probabilities for recommended pathways at each stage by analyzing HCP interactions and prescriptions.
-
The project was selected for poster presentation at the PMSA Conference 2023.
-
We published full paper in the PMSA (Pharmaceutical Management Science Association) Journal Spring 2024.
-
Received Bronze award for this project from Genpact.
2. Assessing the Impact of eSales Aid on Sales using Statistical model :
A. Impact of eSales calls vs. Non eSales calls:
-
Determine how the use of eSales tools during calls affects sales performance compared to traditional calls.
-
Conducted data preparation through segment filtering, selling team categorization, and application of relevant criteria, including data cleaning, outlier detection, and various transformations.
-
Used regression model to quantify the impact of eSales calls on sales, controlling for other variables.
B. Impact of Messages and their combinations :
-
Overall Level: Assess the general effectiveness of different types and sequence of messages on sales.
-
Segments: Evaluate how various customer segments respond to different messages.
-
Cleaned the data, create features for message types and combinations, and segment customers. Explore the data to identify initial patterns and different across segments and specialties.
-
Used regression models to quantify the influence of message types and combinations, providing insights into optimal messaging strategies.
3. Identification of Contribution of Media variables :
-
Conduct market mix modeling to find contribution of media variables and return on investment for reputed clients.'
-
Prepared and analyzed media, non-media and base variables data sets using python.
-
Calculated summary statistics and performed EDA.
-
Employed multiple linear regression models to ascertain the predictive ability and identified crucial areas for targeting.
-
Used different transformations like log-transformation, adstock transformation etc.
-
Calculated return on investment (ROI) for media variables to understand their impact.
Scottish Church College
Here, I joined as a Guest Lecturer for a very short period of time. I took under graduate courses in the area of descriptive statistics and mathematical statistics.
As a research intern I developed a Atmospheric Model using Regression Analysis and prediction algorithm for various atmospheric target variables. Here I used to show the atmospheric response variable by Panoply software which was developed by NASA and the computation method was done in R. I used the supervised machine learning algorithm such as random forest regression for drawing the conclusion. Also in this internship I created Simple White Image Global Warming Poster and SWOT Analysis of Indian Science. The whole process was done under supervision of Prof. Kiran Kumar Namala, IITD.
