Data Scientist
Landmark Group Data Labs · Apr 2022 – Present
- Partnered with BRMs and cross-functional teams (Data Engineering, Platform Engineering, Business Analysts) to translate business requirements into scalable data science solutions, identifying process improvement opportunities through data-driven analysis and Continuous Improvement (CI) methodologies to deliver measurable business value.
- Led regular forecast review sessions with business stakeholders and planners, presenting forecast accuracy results and troubleshooting variances to support customer-facing decision-making and continuous process improvement.
- Built scalable analytical datasets using PySpark, SQL, and Azure Databricks by integrating sales, inventory, promotions, and seasonality data across 10M+ records, enabling feature engineering and improving forecasting accuracy.
- Developed and deployed 16,000+ ML/DL forecasting models monthly across 4,000+ SKUs using regression, tree-based models, gradient boosting (XGBoost), PyTorch-based neural forecasting (NHITS), ARIMA, SARIMA, and Prophet, reducing WMAPE below 20% and lowering stock-outs/overstock by 10% through root-cause variance analysis.
- Applied statistical techniques including exploratory data analysis, hypothesis testing, backtesting, model validation, Design of Experiments (DOE), and hyperparameter tuning to optimize model performance and quantify forecast uncertainty.
- Owned the end-to-end model lifecycle by designing reusable, versioned, and reproducible MLOps pipelines using MLflow, Git, Azure Databricks, and Feature Store for experiment tracking, backtesting, model deployment, performance monitoring, data drift detection, and automated retraining, reducing production runtime by 50%.
- Built Pulse Inventory Insights, an LLM-powered retail analytics solution using PySpark and GPT-based models to generate automated inventory health summaries and planner-friendly business insights across 4,000+ SKUs.

