Data Scientist
ADA GLOBAL · Sep 2023 – Present
- Designed and deployed end-to-end machine learning and GenAI solutions across Databricks and client-managed data platforms for multiple enterprise e-commerce clients, delivering measurable impact across demand forecasting, finance, promotions, and customer analytics.
- Partnered closely with BI, Data Engineering, and client stakeholders to productionize scalable, secure, and reliable data workflows integrated directly into client systems.
- Built a machine learning pipeline to estimate price elasticity for an e-commerce retailer using monthly SKU eligibility checks, automated feature engineering, and data quality validation.
- Automated the entire workflow in Databricks to run monthly and generate SKU-level price elasticity insights.
- Impact: Achieved around 70% accuracy, reduced manual effort, and helped improve pricing decisions to drive better revenue and margins.
- Developed and deployed a deep learning demand forecasting pipeline with automated feature engineering, cost-aware optimization, and Word2Vec product embeddings across Databricks and client ML platforms.
- Impact: Improved forecast accuracy by 20-30% and reduced stockouts by 12-18%.
- Developed a machine learning-driven inventory liquidation planner for an e-commerce retailer to identify products for clearance based on shelf life, Days of Inventory (DOI), expiry, and inventory health.
- Automated the end-to-end pipeline using Databricks to generate periodic liquidation recommendations.
- Impact: Reduced manual planning effort by ~80%, improved inventory turnover by ~20%, and reduced excess inventory carrying costs by ~15%.
- Developed an end-to-end product affinity recommendation pipeline for an e-commerce retailer using the FP-Growth algorithm to identify high-performing product bundles (up to four items) across sellers and campaigns.
- Built automated bi-monthly PySpark pipelines and modular dashboards to generate bundle recommendations and evaluate their effectiveness using metrics such as bundle conversion, repeat purchase rate, and new customer acquisition.
- Impact: Improved bundle conversion by ~15%, and enabled data-driven bundle recommendations to increase cross-sell opportunities and GMV.
- Developed a multi-agent LLM-powered analytics assistant to automate SQL query generation, database execution, root cause analysis, and visualization-based reporting.
- Implemented LangChain-based agents for SQL generation, visualization, and RCA with secure query execution, and built a Streamlit chat interface with inline SQL editing and dynamic charts for interactive data exploration.
- Impact: Reduced dependency on data teams, accelerated insight generation through AI-powered natural language data exploration.


