Bhavya Doshi

Evaluating RAG pipeline architectures with Qdrant for internal knowledge retrieval tooling. • Developing concept frameworks for predictive maintenance and agentic AI integration using LangGraph and MCP based tool orchestration, translating findings into structured product and tooling proposals for the corporate innovation team. • Supporting innovation evaluation through LLM API benchmarking, agent framework assessment, and cross functional workshop facilitation with technical and business stakeholders.

Built SQL queries and Apache Spark based processing workflows for high dimensional relational datasets; improved query execution time by 20% through better extraction and transformation logic. Developed Power BI dashboards to communicate complex research findings to academic stakeholders and support reporting workflows. Co-authored a peer-reviewed Springer publication (2025) on geometrical and regression based modeling, contributing statistical analysis, visualization pipelines, and software development

Automated Salesforce data workflows across B2B and B2C pipelines using Python and dbt transformation logic, cutting manual data latency by 30% and reducing reporting cycle time for 20 commercial stakeholders. Built Power BI dashboards on top of Snowflake-compatible data staging layers for campaign and operational KPIs, enabling faster decision making. Managed structured CRM and sales data across Salesforce and Power BI, supporting commercial reporting for regional B2B operations across the DACH market.

Developed a Python based optimization workflow for automated tape laying of thermoplastic composite laminates, integrating geometry and meshing logic with modular, reproducible pipeline design. Implemented a CLT based screening algorithm to rank stacking sequences under specific load cases and validate structural viability. Validated optimization outputs using simulation tooling, ensuring computational accuracy and reproducibility across engineering test scenarios.

Built Python and PostgreSQL migration scripts for Odoo ERP upgrades, reducing customer query downtime during migration by around 40%. Optimized ERP modules and backend workflows using query tuning and ETL refactoring, improving system functionality and query performance by 25%. Developed automation workflows, API integrations, and responsive JavaScript interfaces to support scalable business operations.

Managed large scale experimental CSV datasets and automated data cleaning pipelines using Python and Pandas to improve prediction accuracy for plasma disruption forecasting. Deployed an LSTM based deep learning model for plasma disruption sequence forecasting, improving prediction reliability across diverse HPC environments. Used high performance computing clusters to optimize computational workflows and support reproducible experimentation.