Apoorva Kulkarni

- Working as part of the analytics team supporting railcar asset and maintenance analytics. My work focuses on improving data reliability and building analytical solutions on large historical datasets used for operational and planning decisions. - Worked on consolidating fragmented historical asset data by developing graph-based logic to connect related records, enabling more consistent analytics and reporting. Built data preprocessing pipelines using Python and SQL to clean, transform, and prepare multi-million record datasets for downstream analysis and modeling. - Contributed to failure prediction initiatives involving large-scale data analysis, extensive exploratory data analysis (EDA), sampling strategies, and training tree-based decision models. The work focused on handling high-volume datasets, feature preparation, and designing decision logic to balance prediction accuracy with operational requirements. - Used Power Automate to automate analytics workflows, improving data refresh reliability and reducing manual reporting effort. - Collaborated across teams to translate analytical outputs into practical solutions for internal data and maintenance analysis. - Tools and technologies used include Python, SQL, Pandas, NumPy, Power BI (DAX), Power Automate, exploratory data analysis, statistical analysis, and tree-based machine learning models.

Teaching Assistant for Visual Analytics, supporting instruction and managing a class of 55+ students under Professor Dr. Gabriel Terejanu.

- Selected as part of a small group of university students to work on a GenAI-focused internal project aimed at improving information access across large internal documentation and knowledge repositories. The work focused on building retrieval-based systems to support frequent business queries and reduce manual search effort. - Designed and implemented a retrieval pipeline using FAISS for vector-based document search and similarity matching. Built multi-turn query handling logic using LangGraph to maintain conversational context and improve response continuity across follow-up queries. - Worked on query preprocessing and prompt structuring to handle ambiguous inputs and improve response consistency. Contributed to evaluation workflows to assess retrieval quality and response relevance across different query types. - Collaborated across cross-functional teams to validate outputs, refine system behavior, and improve usability for internal users. - Tools and technologies used include Python, LangGraph, FAISS, Databricks, retrieval-augmented generation (RAG), prompt engineering, vector embeddings, and evaluation workflows for response quality.

- Worked on backend data extraction and monitoring tasks using Azure-based systems. The role focused on writing queries and building dashboards to support visibility into software performance and operational metrics. - Wrote KQL queries to retrieve and analyze data from Azure data sources used by internal software systems. Built Grafana dashboards using defined KPIs to track performance, reliability, and usage metrics. - Worked with Azure DevOps for task tracking and collaboration as part of Agile development cycles, supporting dashboard updates and monitoring-related improvements. Contributed to development tasks through sprint-based workflows and team collaboration. - Tools and technologies used include KQL, Azure, Azure DevOps, Grafana, data visualization, and Agile (Scrum) practices.