Thejaswi Vepati

Architected and launched a faculty research-submission portal using Spring Boot and React, replacing a manual email-based intake process and reducing administrative processing time by 5 hours per week across two departments. Integrated a conversational support service into an existing Spring Boot application using REST APIs and OpenAI models. Containerized the application with Docker and configured a GitHub Actions CI/CD pipeline, reducing manual deployment time from 2 hours to under 20 minutes. Implemented JWT authentication and role-based access control (RBAC) for student, faculty, and admin roles to secure sensitive research records. Steered sprint execution in bi-weekly reviews with a 4-person team, closing feature tickets and bug reports based on faculty stakeholder feedback.

Deployed new Java/Spring Boot REST API endpoints to expose member-eligibility data to downstream claims-processing services for roughly 8,000 plan members. Refactored a legacy batch job by replacing synchronous database calls with asynchronous Spring Batch processing, reducing nightly runtime from 4 hours to under 90 minutes. Wrote JUnit 5 and Mockito tests that increased unit-test coverage for core service classes from 48% to 82%, uncovering edge-case defects prior to release. Resolved multiple GitHub issues by contributing bug fixes and technical documentation to an internal Java SDK used across three engineering teams.

Designed and delivered a modular Django REST Framework service with PostgreSQL to expose audit data across three client tools, establishing a shared API contract that removed duplicated data-access logic. Separated business logic from ETL jobs by adding a service-layer abstraction over data ingestion, enabling multiple teams to consume audit data through versioned endpoints. Streamlined regulatory report generation using Python and Pandas, cutting report preparation time from 3 days to 4 hours per reporting cycle. Embedded a scikit-learn anomaly-detection model into the audit pipeline to flag unusual transactions and surface results for manual reviewer attention. Collaborated with a cross-functional team using Jira and Confluence while sustaining delivery with zero missed deadlines.