Shubham Gupta

Developed personalized ad-serving solutions using Multi-Armed Bandit (MAB) with Thomson Sampling and Contextual MAB (CMAB) algorithms from ~10k creatives, resulting in a ~20% increase in Click-Through Rate (CTR) and ~1.5x ROAS. Implemented scalable machine learning pipelines with ~50k parameters using PySpark on GCP. Worked with the team in building a testbed simulator in Python for algorithm evaluation using Off-Policy Evaluation (OPE) and Monte Carlo simulations in controlled setting to estimate the expected rewards of new policy.

Created an A/B group creator utility using Python, constraint programming (SMT/SAT solvers), and clustering, it reduced computation time and ensured contamination free experiments.

Data Scientist II - 24/7.ai Labs - Bangalore

(2017-11 - 2024-09)

• Developed personalized ad-serving solutions using Multi-Armed Bandit (MAB) with Thomson Sampling and Contextual MAB (CMAB) algorithms from ~10k creatives, resulting in a ~20% increase in Click-Through Rate (CTR) and ~1.5x ROAS. Implemented scalable machine learning pipelines with ~50k parameters using PySpark on GCP.
• Worked with the team in building a testbed simulator in Python for algorithm evaluation using Off-Policy Evaluation (OPE) and Monte Carlo simulations in controlled setting to estimate the expected rewards of new policy.
• Created an A/B group creator utility using Python, constraint programming (SMT/SAT solvers), and clustering, it reduced computation time and ensured contamination free experiments.
• Automated campaign performance tracking using Python, Seaborn, and Matplotlib. It reduced manual analysis time by ~90%. Performed exploratory data analysis using python libraries.
• Conducted advanced statistical analysis using Cohen's estimator for accurate performance measurement of advertising campaigns. Conducted data visualization using python libraries.
• Implemented purchase propensity models using Logistic Regression and SVM for chatbot interventions to target proactive purchasers, which led us increase sales conversions by ~30% and improving NPS scores by ~20%.
• Worked closely in cross-functional teams with Software Engineers, Data Engineers, Analytics and Product Managers.

Senior Analytics Consultant - Infosys - Bangalore

(2015-02 - 2017-10)

• Employed a sophisticated a mental health chatbot using Naïve Bayes for real-time sentiment analysis. It does dynamic emotion detection from patient utterances through mobile application. Based on recommendation it has capability to reduce health risk by ~30% through proactive intervention.
• Designed an employee attrition prediction model using classification algorithms, it has opportunity to reduce hiring costs by ~20% by enabling targeted retention strategies.
• Built fraud detection models for trade surveillance using behavioural analytics to help business in prioritizing action on suspicious trades, thus improving identification of market manipulation activities. Conducted data visualization using python libraries.

Software Engineer - Fidelity Investments - Bangalore

(2010-08 - 2015-01)

• Built ETL data pipelines using Informatica and automated data reconciliation using shell scripting and SQL query tuning, it reduced processing time by ~40%.
• Integrated a click-to-compute virtual machine application with an OLTP database, it can create virtual machines with customized configuration on the fly. Designed schema, ER (entity relationship) diagram and maintained database for the application.

Post Graduate Program - Business Analytics and Data Science - Great Lakes Institute of Management, Chennai & Illinois Institute of Technology, Chicago (2016-01 - 2016-12)

Bachelor of Technology - Information Technology - Dr. A.P.J. Abdul Kalam Technical University (2005-05 - 2009-05)