Developed an ML model to predict rail track temperature for the next 24 hours by integrating diverse datasets, including meteorological, rail track, geospatial, and historical temperature data. Conducted comprehensive data preprocessing and feature engineering. Deployed the model as a Databricks job, enhancing track maintenance planning and safety, achieving a 30% improvement in mean squared error over a rule-based model.Developed a predictive model to predict rail track defects within 5 days, improving sensitivity by 28% over a rule-based model. Integrated sensor, operational, Geospatial, Historical Defect and environmental data, performed extensive preprocessing and feature engineering, and deployed the model as Databricks Job for early defect detection, enhancing railway safety and maintenance efficiency.

Created a lead scoring system to generate probability scores for targeting high-potential customers and reducing sales efforts. Addressed imbalanced data using diverse sampling techniques and expanded features from 200 to 700 through extensive feature engineering, achieving an AUC score of 0.92. Successfully identified 99% of converted leads in the top 25% of the data, cutting the sales team’s effort by 75%.Implemented Python and PySpark scripts using the Databricks API to retrieve historical stats for jobs and send email notifications to selected users. Utilized the Databricks dashboard for visualizing job statistics, enabling monitoring and analysis of job performance.Developed a system to recommend the most similar stores based on selected stores and analyze pre-time frame and post-time frame sales for both RTM and non-RTM tests across different channels and markets, enabling clients to experiment with various test types on their stores. Enhancing clients decision-making and strategy formulation.

Implemented a machine learning and computer vision algorithm for a virtual interviewing tool, transforming the hiring process and delivering precise interview scores.Developed a deep learning model for short-term wind speed forecasting, achieving a Mean Squared Error of 1.7 km/h and an R² score of 0.88.