At 24/7.ai Labs, I developed an ad-serving solution using Multi-Armed Bandit with Thomson Sampling to address the cold-start problem across ~10k creatives. I implemented a Logistic Regression pipeline with PySpark and GCP that increased CTR by ~20% and delivered ~1.5x ROAS.
I also automated campaign delivery, performance tracking, and optimization through Streamlit, reducing manual effort by ~90% while campaign revenue grew ~8x. I worked with the team on a testbed simulator using Off-Policy Evaluation and Monte Carlo simulations, and created an A/B group utility using constraint programming and clustering.
At Infosys, I worked on a mental health chatbot POC for real-time sentiment analysis and designed an employee attrition prediction model. I also built a fraud-detection visualization utility for trade surveillance using behavioral analytics.
At Fidelity Investments, I built ETL data pipelines with Informatica and automated data reconciliation using shell scripting and SQL query tuning, reducing processing time by ~40%. I also integrated a virtual machine application with an OLTP database and designed and maintained its database schema.
