At Tiger Analytics, I develop end-to-end classification and propensity-scoring pipelines for major retailers, using PySpark and MLflow to predict household shopping frequency tiers.
For a top CPG company, I developed a Databricks pipeline that models buying propensity across 100+ products and generates household-level batch predictions. I also optimized PySpark workflows with partitioning, caching, and query optimization.
I developed an automated metadata-generation framework combining heuristics with Snowflake Cortex LLM capabilities, and implemented a Text-to-SQL solution with Snowflake Cortex Analyst.
Previously, I worked on machine learning, data pipelines, and APIs at Reflections Info Systems PVT LTD, and computer vision and data science projects during internships. My work included rooftop instance segmentation, insurance claim prediction, and a signature and stamp detection model.
