BALAJI CHINTHAGUNTA

Engineered a RAG-based search tool, enabling risk and compliance analysts to query policy and regulatory documents directly, significantly reducing manual review time.

Developed robust document ingestion pipelines, including parsing, chunking, embedding generation, and metadata storage, to facilitate precise filtering by policy type and date.

Implemented advanced LangGraph agent workflows for multi-step queries, integrating internal tools and prioritizing high-risk answers for expedited reviewer attention.

Designed and deployed FastAPI and Spring Boot services with comprehensive input validation and logging, ensuring seamless integration with bank applications and streamlining production issue resolution.

Established a critical prompt versioning system and comprehensive test suite with expected answers, ensuring rigorous validation before any prompt or model changes are deployed.

Fine-tuned small open-source models using SFT and DPO for accurate classification of compliance query intent, deploying them efficiently with VLLM.

Developed and maintained the React frontend for the search and policy Q&A tool, consistently delivering updates through agile sprints in collaboration with product, architecture, and QA teams.

Automated daily claims data preparation using Python and SQL, eliminating a manual spreadsheet process and improving operational efficiency.

Utilized Natural Language Processing (NLP) to classify claim issue types from descriptions and adjuster notes, identifying recurring problems for the claims review team.

Implemented robust SQL checks with joins, CTEs, and aggregations to proactively identify and rectify duplicate claims, missing policy fields, and payment mismatches, ensuring data integrity before reports.

Developed datasets and KPI logic for critical Tableau and Power BI dashboards, supporting weekly and monthly claims reviews and data-driven decision-making.

Contributed to building policyholder segments based on demographics, premium history, and claim frequency, informing renewal and retention analysis strategies.

Created comprehensive documentation for data mappings and validation rules, facilitating seamless pipeline handover and efficient defect resolution for engineering teams.

Built and deployed classification, regression, and forecasting models using scikit-learn, TensorFlow, and PyTorch, accurately predicting maintenance needs and flagging equipment at risk of failure.

Performed extensive data cleaning and validation on engineering, asset, and inspection records using Python and SQL, ensuring high-quality data for model training and analysis.

Leveraged text similarity techniques to group inspection notes, identifying recurring issues and enabling field teams to prioritize critical repairs, improving efficiency.

Conducted rigorous model evaluation using precision, recall, F1, RMSE, and MAE, providing detailed analysis of model performance and limitations to inform selection.

Created interactive Tableau and Power BI dashboards to visualize asset trends and automated weekly and monthly reports, saving significant manual effort and providing timely insights.