Shibaprasad Bhattacharya

Spearhead strategic analytics for clinical trial site selection, reducing timelines from 4 months to under 1 month by analyzing competitive landscapes and enrollment dynamics, leveraging classical statistics (Chi-square, Logistic Regression) and GenAI embedding models to enhance decision quality.

Contribute to the development of an LLM-powered clinical document authoring tool, automating regulatory writing and generating ~$200K in annual savings for the project area, with an enterprise-wide impact of ~$3M.

Develop and deploy an XGBoost-based recommendation model for a specialty pharma product, achieving 79% precision and 37% recall, which reduced marketing overhead by ~18% through improved targeting efficiency.

Analyzed crew operations performance, generating strategic insights and executive dashboards that informed optimal scheduling and resource allocation decisions for leadership.

Designed and deployed a MIP-based Operations Research model, reducing unassigned standby crew by 23% and significantly improving manpower utilization and minimizing operational downtime.

Developed a robust and accurate Bayesian forecasting model to predict incoming load across PAN India facilities, achieving a MAPE score of less than 20 and improving operational planning.

Optimized vendor contract negotiations using mathematical modeling to develop a pricing model, resulting in ~8% cost reduction and enhanced service quality across facility centers.

Rebuilt the last-mile delivery application using Uber H3 hexagons and clustering algorithms, replacing a third-party solver, saving $60K annually, and significantly improving delivery reliability.

Devised a Python-based vehicle recommendation system using the vrpy library, prescribing optimal vehicle types for various centers and enhancing pickup and delivery efficiency.

Optimized and redesigned the service center network using mixed integer programming and open-source solvers, improving efficiency and customer satisfaction, and facilitating the successful merger of Delhivery and SpotOn networks.

Developed a peak load forecasting model using analytical methods and heuristics for high-demand periods, enabling more effective operations planning and resource allocation where traditional ML techniques were insufficient.

Constructed an anomaly detection model for distance reporting using the travelling salesman problem and regression analysis, leading to a significant ~4% reduction in fuel payouts.